Merge pull request #1 from LDOUBLEV/upload

Upload PaddleOCR code 

ppocr/utils/stats.py

+# Copyright (c) 2020 PaddlePaddle Authors. All Rights Reserved.
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+#     http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+
+import collections
+import numpy as np
+import datetime
+
+__all__ = ['TrainingStats', 'Time']
+
+
+class SmoothedValue(object):
+    """Track a series of values and provide access to smoothed values over a
+    window or the global series average.
+    """
+
+    def __init__(self, window_size):
+        self.deque = collections.deque(maxlen=window_size)
+
+    def add_value(self, value):
+        self.deque.append(value)
+
+    def get_median_value(self):
+        return np.median(self.deque)
+
+
+def Time():
+    return datetime.datetime.now().strftime('%Y-%m-%d %H:%M:%S.%f')
+
+
+class TrainingStats(object):
+    def __init__(self, window_size, stats_keys):
+        self.smoothed_losses_and_metrics = {
+            key: SmoothedValue(window_size)
+            for key in stats_keys
+        }
+
+    def update(self, stats):
+        for k, v in self.smoothed_losses_and_metrics.items():
+            v.add_value(stats[k])
+
+    def get(self, extras=None):
+        stats = collections.OrderedDict()
+        if extras:
+            for k, v in extras.items():
+                stats[k] = v
+        for k, v in self.smoothed_losses_and_metrics.items():
+            stats[k] = round(v.get_median_value(), 6)
+
+        return stats
+
+    def log(self, extras=None):
+        d = self.get(extras)
+        strs = ', '.join(str(dict({x: y})).strip('{}') for x, y in d.items())
+        return strs

ppocr/utils/utility.py

+# Copyright (c) 2020 PaddlePaddle Authors. All Rights Reserved.
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+#     http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+
+import logging
+
+
+def initial_logger():
+    FORMAT = '%(asctime)s-%(levelname)s: %(message)s'
+    logging.basicConfig(level=logging.INFO, format=FORMAT)
+    logger = logging.getLogger(__name__)
+    return logger
+
+
+import importlib
+
+
+def create_module(module_str):
+    tmpss = module_str.split(",")
+    assert len(tmpss) == 2, "Error formate\
+        of the module path: {}".format(module_str)
+    module_name, function_name = tmpss[0], tmpss[1]
+    somemodule = importlib.import_module(module_name, __package__)
+    function = getattr(somemodule, function_name)
+    return function
+
+
+def get_check_global_params(mode):
+    check_params = ['use_gpu', 'max_text_length', 'image_shape',\
+        'image_shape', 'character_type', 'loss_type']
+    if mode == "train_eval":
+        check_params = check_params + [\
+            'train_batch_size_per_card', 'test_batch_size_per_card']
+    elif mode == "test":
+        check_params = check_params + ['test_batch_size_per_card']
+    return check_params
+
+
+def get_check_reader_params(mode):
+    check_params = []
+    if mode == "train_eval":
+        check_params = ['TrainReader', 'EvalReader']
+    elif mode == "test":
+        check_params = ['TestReader']
+    return check_params
+
+
+from paddle import fluid
+
+
+def create_multi_devices_program(program, loss_var_name):
+    build_strategy = fluid.BuildStrategy()
+    build_strategy.memory_optimize = False
+    build_strategy.enable_inplace = True
+    exec_strategy = fluid.ExecutionStrategy()
+    exec_strategy.num_iteration_per_drop_scope = 1
+    compile_program = fluid.CompiledProgram(program).with_data_parallel(
+        loss_name=loss_var_name,
+        build_strategy=build_strategy,
+        exec_strategy=exec_strategy)
+    return compile_program

tools/eval.py

+# Copyright (c) 2020 PaddlePaddle Authors. All Rights Reserved.
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+#     http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+
+from __future__ import absolute_import
+from __future__ import division
+from __future__ import print_function
+
+import os
+
+
+def set_paddle_flags(**kwargs):
+    for key, value in kwargs.items():
+        if os.environ.get(key, None) is None:
+            os.environ[key] = str(value)
+
+
+# NOTE(paddle-dev): All of these flags should be
+# set before `import paddle`. Otherwise, it would
+# not take any effect.
+set_paddle_flags(
+    FLAGS_eager_delete_tensor_gb=0,  # enable GC to save memory
+)
+
+import program
+from paddle import fluid
+from ppocr.utils.utility import initial_logger
+logger = initial_logger()
+from ppocr.data.reader_main import reader_main
+from ppocr.utils.save_load import init_model
+from eval_utils.eval_det_utils import eval_det_run
+from eval_utils.eval_rec_utils import test_rec_benchmark
+from eval_utils.eval_rec_utils import eval_rec_run
+from ppocr.utils.character import CharacterOps
+
+
+def main():
+    config = program.load_config(FLAGS.config)
+    program.merge_config(FLAGS.opt)
+    logger.info(config)
+
+    # check if set use_gpu=True in paddlepaddle cpu version
+    use_gpu = config['Global']['use_gpu']
+    program.check_gpu(True)
+
+    alg = config['Global']['algorithm']
+    assert alg in ['EAST', 'DB', 'Rosetta', 'CRNN', 'STARNet', 'RARE']
+    if alg in ['Rosetta', 'CRNN', 'STARNet', 'RARE']:
+        config['Global']['char_ops'] = CharacterOps(config['Global'])
+
+    place = fluid.CUDAPlace(0) if use_gpu else fluid.CPUPlace()
+    startup_prog = fluid.Program()
+    eval_program = fluid.Program()
+    eval_build_outputs = program.build(
+        config, eval_program, startup_prog, mode='test')
+    eval_fetch_name_list = eval_build_outputs[1]
+    eval_fetch_varname_list = eval_build_outputs[2]
+    eval_program = eval_program.clone(for_test=True)
+    exe = fluid.Executor(place)
+    exe.run(startup_prog)
+
+    init_model(config, eval_program, exe)
+
+    if alg in ['EAST', 'DB']:
+        eval_reader = reader_main(config=config, mode="test")
+        eval_info_dict = {'program':eval_program,\
+            'reader':eval_reader,\
+            'fetch_name_list':eval_fetch_name_list,\
+            'fetch_varname_list':eval_fetch_varname_list}
+        metrics = eval_det_run(exe, config, eval_info_dict, "test")
+    else:
+        dataset = config['Global']['dataset']
+        assert dataset in ['lmdb', 'common']
+        if dataset == 'common':
+            eval_reader = reader_main(config=config, mode="eval")
+            eval_info_dict = {'program': eval_program, \
+                              'reader': eval_reader, \
+                              'fetch_name_list': eval_fetch_name_list, \
+                              'fetch_varname_list': eval_fetch_varname_list}
+            metrics = eval_rec_run(exe, config, eval_info_dict, "eval")
+            print("Eval result:", metrics)
+        else:
+            eval_info_dict = {'program':eval_program,\
+                'fetch_name_list':eval_fetch_name_list,\
+                'fetch_varname_list':eval_fetch_varname_list}
+            test_rec_benchmark(exe, config, eval_info_dict)
+
+
+if __name__ == '__main__':
+    parser = program.ArgsParser()
+    FLAGS = parser.parse_args()
+    main()

tools/eval_utils/__init__.py

+# Copyright (c) 2020 PaddlePaddle Authors. All Rights Reserved.
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+#     http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.

tools/eval_utils/eval_det_iou.py

+#!/usr/bin/env python
+# -*- coding: utf-8 -*-
+from collections import namedtuple
+import numpy as np
+from shapely.geometry import Polygon
+
+
+class DetectionIoUEvaluator(object):
+    def __init__(self, iou_constraint=0.5, area_precision_constraint=0.5):
+        self.iou_constraint = iou_constraint
+        self.area_precision_constraint = area_precision_constraint
+
+    def evaluate_image(self, gt, pred):
+        def get_union(pD, pG):
+            return Polygon(pD).union(Polygon(pG)).area
+
+        def get_intersection_over_union(pD, pG):
+            return get_intersection(pD, pG) / get_union(pD, pG)
+
+        def get_intersection(pD, pG):
+            return Polygon(pD).intersection(Polygon(pG)).area
+
+        def compute_ap(confList, matchList, numGtCare):
+            correct = 0
+            AP = 0
+            if len(confList) > 0:
+                confList = np.array(confList)
+                matchList = np.array(matchList)
+                sorted_ind = np.argsort(-confList)
+                confList = confList[sorted_ind]
+                matchList = matchList[sorted_ind]
+                for n in range(len(confList)):
+                    match = matchList[n]
+                    if match:
+                        correct += 1
+                        AP += float(correct) / (n + 1)
+
+                if numGtCare > 0:
+                    AP /= numGtCare
+
+            return AP
+
+        perSampleMetrics = {}
+
+        matchedSum = 0
+
+        Rectangle = namedtuple('Rectangle', 'xmin ymin xmax ymax')
+
+        numGlobalCareGt = 0
+        numGlobalCareDet = 0
+
+        arrGlobalConfidences = []
+        arrGlobalMatches = []
+
+        recall = 0
+        precision = 0
+        hmean = 0
+
+        detMatched = 0
+
+        iouMat = np.empty([1, 1])
+
+        gtPols = []
+        detPols = []
+
+        gtPolPoints = []
+        detPolPoints = []
+
+        # Array of Ground Truth Polygons' keys marked as don't Care
+        gtDontCarePolsNum = []
+        # Array of Detected Polygons' matched with a don't Care GT
+        detDontCarePolsNum = []
+
+        pairs = []
+        detMatchedNums = []
+
+        arrSampleConfidences = []
+        arrSampleMatch = []
+
+        evaluationLog = ""
+
+        # print(len(gt))
+        for n in range(len(gt)):
+            points = gt[n]['points']
+            # transcription = gt[n]['text']
+            dontCare = gt[n]['ignore']
+            points = Polygon(points)
+            points = points.buffer(0)
+            if not Polygon(points).is_valid or not Polygon(points).is_simple:
+                continue
+
+            gtPol = points
+            gtPols.append(gtPol)
+            gtPolPoints.append(points)
+            if dontCare:
+                gtDontCarePolsNum.append(len(gtPols) - 1)
+
+        evaluationLog += "GT polygons: " + str(len(gtPols)) + (
+            " (" + str(len(gtDontCarePolsNum)) + " don't care)\n"
+            if len(gtDontCarePolsNum) > 0 else "\n")
+
+        for n in range(len(pred)):
+            points = pred[n]['points']
+            points = Polygon(points)
+            points = points.buffer(0)
+            if not Polygon(points).is_valid or not Polygon(points).is_simple:
+                continue
+
+            detPol = points
+            detPols.append(detPol)
+            detPolPoints.append(points)
+            if len(gtDontCarePolsNum) > 0:
+                for dontCarePol in gtDontCarePolsNum:
+                    dontCarePol = gtPols[dontCarePol]
+                    intersected_area = get_intersection(dontCarePol, detPol)
+                    pdDimensions = Polygon(detPol).area
+                    precision = 0 if pdDimensions == 0 else intersected_area / pdDimensions
+                    if (precision > self.area_precision_constraint):
+                        detDontCarePolsNum.append(len(detPols) - 1)
+                        break
+
+        evaluationLog += "DET polygons: " + str(len(detPols)) + (
+            " (" + str(len(detDontCarePolsNum)) + " don't care)\n"
+            if len(detDontCarePolsNum) > 0 else "\n")
+
+        if len(gtPols) > 0 and len(detPols) > 0:
+            # Calculate IoU and precision matrixs
+            outputShape = [len(gtPols), len(detPols)]
+            iouMat = np.empty(outputShape)
+            gtRectMat = np.zeros(len(gtPols), np.int8)
+            detRectMat = np.zeros(len(detPols), np.int8)
+            for gtNum in range(len(gtPols)):
+                for detNum in range(len(detPols)):
+                    pG = gtPols[gtNum]
+                    pD = detPols[detNum]
+                    iouMat[gtNum, detNum] = get_intersection_over_union(pD, pG)
+
+            for gtNum in range(len(gtPols)):
+                for detNum in range(len(detPols)):
+                    if gtRectMat[gtNum] == 0 and detRectMat[
+                            detNum] == 0 and gtNum not in gtDontCarePolsNum and detNum not in detDontCarePolsNum:
+                        if iouMat[gtNum, detNum] > self.iou_constraint:
+                            gtRectMat[gtNum] = 1
+                            detRectMat[detNum] = 1
+                            detMatched += 1
+                            pairs.append({'gt': gtNum, 'det': detNum})
+                            detMatchedNums.append(detNum)
+                            evaluationLog += "Match GT #" + \
+                                str(gtNum) + " with Det #" + str(detNum) + "\n"
+
+        numGtCare = (len(gtPols) - len(gtDontCarePolsNum))
+        numDetCare = (len(detPols) - len(detDontCarePolsNum))
+        if numGtCare == 0:
+            recall = float(1)
+            precision = float(0) if numDetCare > 0 else float(1)
+        else:
+            recall = float(detMatched) / numGtCare
+            precision = 0 if numDetCare == 0 else float(detMatched) / numDetCare
+
+        hmean = 0 if (precision + recall) == 0 else 2.0 * \
+            precision * recall / (precision + recall)
+
+        matchedSum += detMatched
+        numGlobalCareGt += numGtCare
+        numGlobalCareDet += numDetCare
+
+        perSampleMetrics = {
+            'precision': precision,
+            'recall': recall,
+            'hmean': hmean,
+            'pairs': pairs,
+            'iouMat': [] if len(detPols) > 100 else iouMat.tolist(),
+            'gtPolPoints': gtPolPoints,
+            'detPolPoints': detPolPoints,
+            'gtCare': numGtCare,
+            'detCare': numDetCare,
+            'gtDontCare': gtDontCarePolsNum,
+            'detDontCare': detDontCarePolsNum,
+            'detMatched': detMatched,
+            'evaluationLog': evaluationLog
+        }
+
+        return perSampleMetrics
+
+    def combine_results(self, results):
+        numGlobalCareGt = 0
+        numGlobalCareDet = 0
+        matchedSum = 0
+        for result in results:
+            numGlobalCareGt += result['gtCare']
+            numGlobalCareDet += result['detCare']
+            matchedSum += result['detMatched']
+
+        methodRecall = 0 if numGlobalCareGt == 0 else float(
+            matchedSum) / numGlobalCareGt
+        methodPrecision = 0 if numGlobalCareDet == 0 else float(
+            matchedSum) / numGlobalCareDet
+        methodHmean = 0 if methodRecall + methodPrecision == 0 else 2 * \
+            methodRecall * methodPrecision / (methodRecall + methodPrecision)
+        # print(methodRecall, methodPrecision, methodHmean)
+        # sys.exit(-1)
+        methodMetrics = {
+            'precision': methodPrecision,
+            'recall': methodRecall,
+            'hmean': methodHmean
+        }
+
+        return methodMetrics
+
+
+if __name__ == '__main__':
+    evaluator = DetectionIoUEvaluator()
+    gts = [[{
+        'points': [(0, 0), (1, 0), (1, 1), (0, 1)],
+        'text': 1234,
+        'ignore': False,
+    }, {
+        'points': [(2, 2), (3, 2), (3, 3), (2, 3)],
+        'text': 5678,
+        'ignore': False,
+    }]]
+    preds = [[{
+        'points': [(0.1, 0.1), (1, 0), (1, 1), (0, 1)],
+        'text': 123,
+        'ignore': False,
+    }]]
+    results = []
+    for gt, pred in zip(gts, preds):
+        results.append(evaluator.evaluate_image(gt, pred))
+    metrics = evaluator.combine_results(results)
+    print(metrics)

tools/eval_utils/eval_det_utils.py

+# Copyright (c) 2020 PaddlePaddle Authors. All Rights Reserved.
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+#     http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+
+from __future__ import absolute_import
+from __future__ import division
+from __future__ import print_function
+
+import logging
+import numpy as np
+
+import paddle.fluid as fluid
+
+__all__ = ['eval_det_run']
+
+import logging
+FORMAT = '%(asctime)s-%(levelname)s: %(message)s'
+logging.basicConfig(level=logging.INFO, format=FORMAT)
+logger = logging.getLogger(__name__)
+
+from ppocr.utils.utility import create_module
+from .eval_det_iou import DetectionIoUEvaluator
+import json
+from copy import deepcopy
+import cv2
+from ppocr.data.reader_main import reader_main
+
+
+def cal_det_res(exe, config, eval_info_dict):
+    global_params = config['Global']
+    save_res_path = global_params['save_res_path']
+    postprocess_params = deepcopy(config["PostProcess"])
+    postprocess_params.update(global_params)
+    postprocess = create_module(postprocess_params['function']) \
+        (params=postprocess_params)
+    with open(save_res_path, "wb") as fout:
+        tackling_num = 0
+        for data in eval_info_dict['reader']():
+            img_num = len(data)
+            tackling_num = tackling_num + img_num
+            logger.info("test tackling num:%d", tackling_num)
+            img_list = []
+            ratio_list = []
+            img_name_list = []
+            for ino in range(img_num):
+                img_list.append(data[ino][0])
+                ratio_list.append(data[ino][1])
+                img_name_list.append(data[ino][2])
+            img_list = np.concatenate(img_list, axis=0)
+            outs = exe.run(eval_info_dict['program'], \
+                           feed={'image': img_list}, \
+                           fetch_list=eval_info_dict['fetch_varname_list'])
+            outs_dict = {}
+            for tno in range(len(outs)):
+                fetch_name = eval_info_dict['fetch_name_list'][tno]
+                fetch_value = np.array(outs[tno])
+                outs_dict[fetch_name] = fetch_value
+            dt_boxes_list = postprocess(outs_dict, ratio_list)
+            for ino in range(img_num):
+                dt_boxes = dt_boxes_list[ino]
+                img_name = img_name_list[ino]
+                dt_boxes_json = []
+                for box in dt_boxes:
+                    tmp_json = {"transcription": ""}
+                    tmp_json['points'] = box.tolist()
+                    dt_boxes_json.append(tmp_json)
+                otstr = img_name + "\t" + json.dumps(dt_boxes_json) + "\n"
+                fout.write(otstr.encode())
+    return
+
+
+def load_label_infor(label_file_path, do_ignore=False):
+    img_name_label_dict = {}
+    with open(label_file_path, "rb") as fin:
+        lines = fin.readlines()
+        for line in lines:
+            substr = line.decode().strip("\n").split("\t")
+            bbox_infor = json.loads(substr[1])
+            bbox_num = len(bbox_infor)
+            for bno in range(bbox_num):
+                text = bbox_infor[bno]['transcription']
+                ignore = False
+                if text == "###" and do_ignore:
+                    ignore = True
+                bbox_infor[bno]['ignore'] = ignore
+            img_name_label_dict[substr[0]] = bbox_infor
+    return img_name_label_dict
+
+
+def cal_det_metrics(gt_label_path, save_res_path):
+    evaluator = DetectionIoUEvaluator()
+    gt_label_infor = load_label_infor(gt_label_path, do_ignore=True)
+    dt_label_infor = load_label_infor(save_res_path)
+    results = []
+    for img_name in gt_label_infor:
+        gt_label = gt_label_infor[img_name]
+        if img_name not in dt_label_infor:
+            dt_label = []
+        else:
+            dt_label = dt_label_infor[img_name]
+        result = evaluator.evaluate_image(gt_label, dt_label)
+        results.append(result)
+    methodMetrics = evaluator.combine_results(results)
+    return methodMetrics
+
+
+def eval_det_run(exe, config, eval_info_dict, mode):
+    cal_det_res(exe, config, eval_info_dict)
+
+    save_res_path = config['Global']['save_res_path']
+    if mode == "eval":
+        gt_label_path = config['EvalReader']['label_file_path']
+        metrics = cal_det_metrics(gt_label_path, save_res_path)
+    else:
+        gt_label_path = config['TestReader']['label_file_path']
+        do_eval = config['TestReader']['do_eval']
+        if do_eval:
+            metrics = cal_det_metrics(gt_label_path, save_res_path)
+        else:
+            metrics = {}
+    return metrics

tools/eval_utils/eval_rec_utils.py

+# Copyright (c) 2020 PaddlePaddle Authors. All Rights Reserved.
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+#     http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+
+from __future__ import absolute_import
+from __future__ import division
+from __future__ import print_function
+
+import logging
+import numpy as np
+
+import paddle.fluid as fluid
+
+__all__ = ['eval_rec_run', 'test_rec_benchmark']
+
+import logging
+
+FORMAT = '%(asctime)s-%(levelname)s: %(message)s'
+logging.basicConfig(level=logging.INFO, format=FORMAT)
+logger = logging.getLogger(__name__)
+
+from ppocr.utils.character import cal_predicts_accuracy
+from ppocr.utils.character import convert_rec_label_to_lod
+from ppocr.utils.character import convert_rec_attention_infer_res
+from ppocr.utils.utility import create_module
+import json
+from copy import deepcopy
+import cv2
+from ppocr.data.reader_main import reader_main
+
+
+def eval_rec_run(exe, config, eval_info_dict, mode):
+    """
+    Run evaluation program, return program outputs.
+    """
+    char_ops = config['Global']['char_ops']
+    total_loss = 0
+    total_sample_num = 0
+    total_acc_num = 0
+    total_batch_num = 0
+    if mode == "eval":
+        is_remove_duplicate = False
+    else:
+        is_remove_duplicate = True
+
+    for data in eval_info_dict['reader']():
+        img_num = len(data)
+        img_list = []
+        label_list = []
+        for ino in range(img_num):
+            img_list.append(data[ino][0])
+            label_list.append(data[ino][1])
+        img_list = np.concatenate(img_list, axis=0)
+        outs = exe.run(eval_info_dict['program'], \
+                       feed={'image': img_list}, \
+                       fetch_list=eval_info_dict['fetch_varname_list'], \
+                       return_numpy=False)
+        preds = np.array(outs[0])
+        if preds.shape[1] != 1:
+            preds, preds_lod = convert_rec_attention_infer_res(preds)
+        else:
+            preds_lod = outs[0].lod()[0]
+        labels, labels_lod = convert_rec_label_to_lod(label_list)
+        acc, acc_num, sample_num = cal_predicts_accuracy(
+            char_ops, preds, preds_lod, labels, labels_lod, is_remove_duplicate)
+        total_acc_num += acc_num
+        total_sample_num += sample_num
+        total_batch_num += 1
+    avg_acc = total_acc_num * 1.0 / total_sample_num
+    metrics = {'avg_acc': avg_acc, "total_acc_num": total_acc_num, \
+               "total_sample_num": total_sample_num}
+    return metrics
+
+
+def test_rec_benchmark(exe, config, eval_info_dict):
+    " 评估lmdb 数据"
+    eval_data_list = ['IIIT5k_3000', 'SVT', 'IC03_860', 'IC03_867', \
+                      'IC13_857', 'IC13_1015', 'IC15_1811', 'IC15_2077', 'SVTP', 'CUTE80']
+    eval_data_dir = config['TestReader']['lmdb_sets_dir']
+    total_evaluation_data_number = 0
+    total_correct_number = 0
+    eval_data_acc_info = {}
+    for eval_data in eval_data_list:
+        config['TestReader']['lmdb_sets_dir'] = \
+            eval_data_dir + "/" + eval_data
+        eval_reader = reader_main(config=config, mode="test")
+        eval_info_dict['reader'] = eval_reader
+        metrics = eval_rec_run(exe, config, eval_info_dict, "test")
+        total_evaluation_data_number += metrics['total_sample_num']
+        total_correct_number += metrics['total_acc_num']
+        eval_data_acc_info[eval_data] = metrics
+
+    avg_acc = total_correct_number * 1.0 / total_evaluation_data_number
+    logger.info('-' * 50)
+    strs = ""
+    for eval_data in eval_data_list:
+        eval_acc = eval_data_acc_info[eval_data]['avg_acc']
+        strs += "\n {}, accuracy:{:.6f}".format(eval_data, eval_acc)
+    strs += "\n average, accuracy:{:.6f}".format(avg_acc)
+    logger.info(strs)
+    logger.info('-' * 50)

tools/export_model.py

+# Copyright (c) 2020 PaddlePaddle Authors. All Rights Reserved.
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+#     http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+
+from __future__ import absolute_import
+from __future__ import division
+from __future__ import print_function
+
+import os
+import sys
+import time
+import multiprocessing
+import numpy as np
+
+
+def set_paddle_flags(**kwargs):
+    for key, value in kwargs.items():
+        if os.environ.get(key, None) is None:
+            os.environ[key] = str(value)
+
+
+# NOTE(paddle-dev): All of these flags should be
+# set before `import paddle`. Otherwise, it would
+# not take any effect. 
+set_paddle_flags(
+    FLAGS_eager_delete_tensor_gb=0,  # enable GC to save memory
+)
+
+import program
+from paddle import fluid
+from ppocr.utils.utility import initial_logger
+logger = initial_logger()
+from ppocr.utils.save_load import init_model
+from ppocr.utils.character import CharacterOps
+from ppocr.utils.utility import create_module
+
+
+def main():
+    config = program.load_config(FLAGS.config)
+    program.merge_config(FLAGS.opt)
+    logger.info(config)
+
+    # check if set use_gpu=True in paddlepaddle cpu version
+    use_gpu = config['Global']['use_gpu']
+    program.check_gpu(True)
+
+    alg = config['Global']['algorithm']
+    assert alg in ['EAST', 'DB', 'Rosetta', 'CRNN', 'STARNet', 'RARE']
+    if alg in ['Rosetta', 'CRNN', 'STARNet', 'RARE']:
+        config['Global']['char_ops'] = CharacterOps(config['Global'])
+
+    place = fluid.CUDAPlace(0) if use_gpu else fluid.CPUPlace()
+    startup_prog = fluid.Program()
+    eval_program = fluid.Program()
+
+    feeded_var_names, target_vars, fetches_var_name = program.build_export(
+        config, eval_program, startup_prog)
+    eval_program = eval_program.clone(for_test=True)
+    exe = fluid.Executor(place)
+    exe.run(startup_prog)
+
+    init_model(config, eval_program, exe)
+
+    fluid.io.save_inference_model(
+        dirname="./output/",
+        feeded_var_names=feeded_var_names,
+        main_program=eval_program,
+        target_vars=target_vars,
+        executor=exe,
+        model_filename='model',
+        params_filename='params')
+    print("save success, output_name_list:", fetches_var_name)
+
+
+if __name__ == '__main__':
+    parser = program.ArgsParser()
+    FLAGS = parser.parse_args()
+    main()

tools/infer/det_program.txt

+<paddle.fluid.core_avx.ProgramDesc object at 0x10d15fab0>

tools/infer/predict_det.py

+# Copyright (c) 2020 PaddlePaddle Authors. All Rights Reserved.
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+#     http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+
+import utility
+from ppocr.utils.utility import initial_logger
+logger = initial_logger()
+import cv2
+from ppocr.data.det.east_process import EASTProcessTest
+from ppocr.data.det.db_process import DBProcessTest
+from ppocr.postprocess.db_postprocess import DBPostProcess
+from ppocr.postprocess.east_postprocess import EASTPostPocess
+import copy
+import numpy as np
+import math
+import time
+
+
+class TextDetector(object):
+    def __init__(self, args):
+        max_side_len = args.det_max_side_len
+        self.det_algorithm = args.det_algorithm
+        preprocess_params = {'max_side_len': max_side_len}
+        postprocess_params = {}
+        if self.det_algorithm == "DB":
+            self.preprocess_op = DBProcessTest(preprocess_params)
+            postprocess_params["thresh"] = args.det_db_thresh
+            postprocess_params["box_thresh"] = args.det_db_box_thresh
+            postprocess_params["max_candidates"] = 1000
+            self.postprocess_op = DBPostProcess(postprocess_params)
+        elif self.det_algorithm == "EAST":
+            self.preprocess_op = EASTProcessTest(preprocess_params)
+            postprocess_params["score_thresh"] = args.det_east_score_thresh
+            postprocess_params["cover_thresh"] = args.det_east_cover_thresh
+            postprocess_params["nms_thresh"] = args.det_east_nms_thresh
+            self.postprocess_op = EASTPostPocess(postprocess_params)
+        else:
+            logger.info("unknown det_algorithm:{}".format(self.det_algorithm))
+            sys.exit(0)
+
+        self.predictor, self.input_tensor, self.output_tensors =\
+            utility.create_predictor(args, mode="det")
+
+    def order_points_clockwise(self, pts):
+        #######
+        ## https://github.com/jrosebr1/imutils/blob/master/imutils/perspective.py
+        ########
+        # sort the points based on their x-coordinates
+        xSorted = pts[np.argsort(pts[:, 0]), :]
+
+        # grab the left-most and right-most points from the sorted
+        # x-roodinate points
+        leftMost = xSorted[:2, :]
+        rightMost = xSorted[2:, :]
+
+        # now, sort the left-most coordinates according to their
+        # y-coordinates so we can grab the top-left and bottom-left
+        # points, respectively
+        leftMost = leftMost[np.argsort(leftMost[:, 1]), :]
+        (tl, bl) = leftMost
+
+        rightMost = rightMost[np.argsort(rightMost[:, 1]), :]
+        (tr, br) = rightMost
+
+        rect = np.array([tl, tr, br, bl], dtype="float32")
+        return rect
+
+    def expand_det_res(self, points, bbox_height, bbox_width, img_height,
+                       img_width):
+        if bbox_height * 1.0 / bbox_width >= 2.0:
+            expand_w = bbox_width * 0.20
+            expand_h = bbox_width * 0.20
+        elif bbox_width * 1.0 / bbox_height >= 3.0:
+            expand_w = bbox_height * 0.20
+            expand_h = bbox_height * 0.20
+        else:
+            expand_w = bbox_height * 0.1
+            expand_h = bbox_height * 0.1
+
+        points[0, 0] = int(max((points[0, 0] - expand_w), 0))
+        points[1, 0] = int(min((points[1, 0] + expand_w), img_width))
+        points[3, 0] = int(max((points[3, 0] - expand_w), 0))
+        points[2, 0] = int(min((points[2, 0] + expand_w), img_width))
+
+        points[0, 1] = int(max((points[0, 1] - expand_h), 0))
+        points[1, 1] = int(max((points[1, 1] - expand_h), 0))
+        points[3, 1] = int(min((points[3, 1] + expand_h), img_height))
+        points[2, 1] = int(min((points[2, 1] + expand_h), img_height))
+        return points
+
+    def filter_tag_det_res(self, dt_boxes, image_shape):
+        img_height, img_width = image_shape[0:2]
+        dt_boxes_new = []
+        for box in dt_boxes:
+            box = self.order_points_clockwise(box)
+            left = int(np.min(box[:, 0]))
+            right = int(np.max(box[:, 0]))
+            top = int(np.min(box[:, 1]))
+            bottom = int(np.max(box[:, 1]))
+            bbox_height = bottom - top
+            bbox_width = right - left
+            diffh = math.fabs(box[0, 1] - box[1, 1])
+            diffw = math.fabs(box[0, 0] - box[3, 0])
+            rect_width = int(np.linalg.norm(box[0] - box[1]))
+            rect_height = int(np.linalg.norm(box[0] - box[3]))
+            if rect_width <= 10 or rect_height <= 10:
+                continue
+            if diffh <= 10 and diffw <= 10:
+                box = self.expand_det_res(
+                    copy.deepcopy(box), bbox_height, bbox_width, img_height,
+                    img_width)
+            dt_boxes_new.append(box)
+        dt_boxes = np.array(dt_boxes_new)
+        return dt_boxes
+
+    def __call__(self, img):
+        ori_im = img.copy()
+        im, ratio_list = self.preprocess_op(img)
+        if im is None:
+            return None, 0
+        im = im.copy()
+        starttime = time.time()
+        self.input_tensor.copy_from_cpu(im)
+        self.predictor.zero_copy_run()
+        outputs = []
+        for output_tensor in self.output_tensors:
+            output = output_tensor.copy_to_cpu()
+            outputs.append(output)
+        outs_dict = {}
+        if self.det_algorithm == "EAST":
+            outs_dict['f_score'] = outputs[0]
+            outs_dict['f_geo'] = outputs[1]
+        else:
+            outs_dict['maps'] = [outputs[0]]
+        dt_boxes_list = self.postprocess_op(outs_dict, [ratio_list])
+        dt_boxes = dt_boxes_list[0]
+        dt_boxes = self.filter_tag_det_res(dt_boxes, ori_im.shape)
+        elapse = time.time() - starttime
+        return dt_boxes, elapse
+
+
+if __name__ == "__main__":
+    args = utility.parse_args()
+    image_file_list = utility.get_image_file_list(args.image_dir)
+    text_detector = TextDetector(args)
+    count = 0
+    total_time = 0
+    for image_file in image_file_list:
+        img = cv2.imread(image_file)
+        if img is None:
+            logger.info("error in loading image:{}".format(image_file))
+            continue
+        dt_boxes, elapse = text_detector(img)
+        if count > 0:
+            total_time += elapse
+        count += 1
+        print("Predict time of %s:" % image_file, elapse)
+        utility.draw_text_det_res(dt_boxes, image_file)
+    print("Avg Time:", total_time / (count - 1))

tools/infer/predict_eval.py

+# Copyright (c) 2020 PaddlePaddle Authors. All Rights Reserved.
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+#     http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+
+import utility
+from ppocr.utils.utility import initial_logger
+logger = initial_logger()
+import cv2
+import predict_system
+import copy
+import numpy as np
+import math
+import time
+import json
+
+if __name__ == "__main__":
+    args = utility.parse_args()
+    text_sys = predict_system.TextSystem(args)
+
+    image_file_list = []
+    label_file_path = "./eval_perform/gt_res/test_chinese_ic15_500_4pts.txt"
+    img_set_path = "./eval_perform/"
+    with open(label_file_path, "rb") as fin:
+        lines = fin.readlines()
+        for line in lines:
+            substr = line.decode('utf-8').strip("\n").split("\t")
+            if "lsvt" in substr[0]:
+                continue
+            image_file_list.append(substr[0])
+
+    total_time_all = 0
+    count = 0
+    save_path = "./output/predict.txt"
+    fout = open(save_path, "wb")
+    for image_name in image_file_list:
+        image_file = img_set_path + image_name
+        img = cv2.imread(image_file)
+        if img is None:
+            logger.info("error in loading image:{}".format(image_file))
+            continue
+        count += 1
+        total_time = 0
+        starttime = time.time()
+        dt_boxes, rec_res = text_sys(img)
+        elapse = time.time() - starttime
+        total_time_all += elapse
+        print("Predict time of %s(%d): %.3fs" % (image_file, count, elapse))
+        dt_num = len(dt_boxes)
+        bbox_list = []
+        for dno in range(dt_num):
+            box = dt_boxes[dno]
+            text, score = rec_res[dno]
+            points = []
+            for tno in range(len(box)):
+                points.append([box[tno][0] * 1.0, box[tno][1] * 1.0])
+            bbox_list.append({
+                "transcription": text,
+                "points": points,
+                "scores": score * 1.0
+            })
+        otstr = image_name + "\t" + json.dumps(bbox_list) + "\n"
+        fout.write(otstr.encode('utf-8'))
+    avg_time = total_time_all / count
+    logger.info("avg_time: {0}".format(avg_time))
+    logger.info("avg_fps: {0}".format(1.0 / avg_time))
+    fout.close()

tools/infer/predict_eval_new.py

+# Copyright (c) 2020 PaddlePaddle Authors. All Rights Reserved.
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+#     http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+
+import utility
+from ppocr.utils.utility import initial_logger
+logger = initial_logger()
+import cv2
+import predict_system
+import copy
+import numpy as np
+import math
+import time
+import json
+import os
+
+if __name__ == "__main__":
+    args = utility.parse_args()
+    text_sys = predict_system.TextSystem(args)
+
+    image_file_list = []
+    img_set_path = "/paddle/code/dyn/test_imgs/rctw_samples/"
+    image_file_list = os.listdir(img_set_path)
+
+    total_time_all = 0
+    count = 0
+    save_path = "./output/predict.txt"
+    fout = open(save_path, "wb")
+    for image_name in image_file_list:
+        image_file = img_set_path + image_name
+        img = cv2.imread(image_file)
+        if img is None:
+            logger.info("error in loading image:{}".format(image_file))
+            continue
+        count += 1
+        starttime = time.time()
+        dt_boxes, rec_res = text_sys(img)
+        if dt_boxes is None:
+            count -= 1
+            continue
+        elapse = time.time() - starttime
+        total_time_all += elapse
+        print("Predict time of %s(%d): %.3fs" % (image_file, count, elapse))
+        dt_num = len(dt_boxes)
+        bbox_list = []
+        for dno in range(dt_num):
+            box = dt_boxes[dno]
+            text, score = rec_res[dno]
+            points = []
+            for tno in range(len(box)):
+                points.append([box[tno][0] * 1.0, box[tno][1] * 1.0])
+            bbox_list.append({
+                "transcription": text,
+                "points": points,
+                "scores": score * 1.0
+            })
+        otstr = image_name + "\t" + json.dumps(bbox_list) + "\n"
+        fout.write(otstr.encode('utf-8'))
+    avg_time = total_time_all / count
+    logger.info("avg_time: {0}".format(avg_time))
+    logger.info("avg_fps: {0}".format(1.0 / avg_time))
+    fout.close()

tools/infer/predict_rec.py

+# Copyright (c) 2020 PaddlePaddle Authors. All Rights Reserved.
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+#     http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+
+import utility
+from ppocr.utils.utility import initial_logger
+logger = initial_logger()
+import cv2
+
+import copy
+import numpy as np
+import math
+import time
+from ppocr.utils.character import CharacterOps
+
+
+class TextRecognizer(object):
+    def __init__(self, args):
+        self.predictor, self.input_tensor, self.output_tensors =\
+            utility.create_predictor(args, mode="rec")
+        image_shape = [int(v) for v in args.rec_image_shape.split(",")]
+        self.rec_image_shape = image_shape
+        char_ops_params = {}
+        char_ops_params["character_type"] = args.rec_char_type
+        char_ops_params["character_dict_path"] = args.rec_char_dict_path
+        char_ops_params['loss_type'] = 'ctc'
+        self.char_ops = CharacterOps(char_ops_params)
+
+    def resize_norm_img(self, img):
+        imgC, imgH, imgW = self.rec_image_shape
+        h = img.shape[0]
+        w = img.shape[1]
+        ratio = w / float(h)
+        if math.ceil(imgH * ratio) > imgW:
+            resized_w = imgW
+        else:
+            resized_w = int(math.ceil(imgH * ratio))
+        resized_image = cv2.resize(img, (resized_w, imgH))
+        resized_image = resized_image.astype('float32')
+        resized_image = resized_image.transpose((2, 0, 1)) / 255
+        resized_image -= 0.5
+        resized_image /= 0.5
+        padding_im = np.zeros((imgC, imgH, imgW), dtype=np.float32)
+        padding_im[:, :, 0:resized_w] = resized_image
+        return padding_im
+
+    def __call__(self, img_list):
+        img_num = len(img_list)
+        batch_num = 15
+        rec_res = []
+        predict_time = 0
+        for beg_img_no in range(0, img_num, batch_num):
+            end_img_no = min(img_num, beg_img_no + batch_num)
+            norm_img_batch = []
+            for ino in range(beg_img_no, end_img_no):
+                norm_img = self.resize_norm_img(img_list[ino])
+                norm_img = norm_img[np.newaxis, :]
+                norm_img_batch.append(norm_img)
+            norm_img_batch = np.concatenate(norm_img_batch)
+            norm_img_batch = norm_img_batch.copy()
+            starttime = time.time()
+            self.input_tensor.copy_from_cpu(norm_img_batch)
+            self.predictor.zero_copy_run()
+            rec_idx_batch = self.output_tensors[0].copy_to_cpu()
+            rec_idx_lod = self.output_tensors[0].lod()[0]
+            predict_batch = self.output_tensors[1].copy_to_cpu()
+            predict_lod = self.output_tensors[1].lod()[0]
+            elapse = time.time() - starttime
+            predict_time += elapse
+            starttime = time.time()
+            for rno in range(len(rec_idx_lod) - 1):
+                beg = rec_idx_lod[rno]
+                end = rec_idx_lod[rno + 1]
+                rec_idx_tmp = rec_idx_batch[beg:end, 0]
+                preds_text = self.char_ops.decode(rec_idx_tmp)
+                beg = predict_lod[rno]
+                end = predict_lod[rno + 1]
+                probs = predict_batch[beg:end, :]
+                ind = np.argmax(probs, axis=1)
+                blank = probs.shape[1]
+                valid_ind = np.where(ind != (blank - 1))[0]
+                score = np.mean(probs[valid_ind, ind[valid_ind]])
+                rec_res.append([preds_text, score])
+        return rec_res, predict_time
+
+
+if __name__ == "__main__":
+    args = utility.parse_args()
+    image_file_list = utility.get_image_file_list(args.image_dir)
+    text_recognizer = TextRecognizer(args)
+    valid_image_file_list = []
+    img_list = []
+    for image_file in image_file_list:
+        img = cv2.imread(image_file)
+        if img is None:
+            logger.info("error in loading image:{}".format(image_file))
+            continue
+        valid_image_file_list.append(image_file)
+        img_list.append(img)
+    rec_res, predict_time = text_recognizer(img_list)
+    rec_res, predict_time = text_recognizer(img_list)
+    for ino in range(len(img_list)):
+        print("Predicts of %s:%s" % (valid_image_file_list[ino], rec_res[ino]))
+    print("Total predict time for %d images:%.3f" %
+          (len(img_list), predict_time))

tools/infer/predict_system.py

+# Copyright (c) 2020 PaddlePaddle Authors. All Rights Reserved.
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+#     http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+
+import utility
+from ppocr.utils.utility import initial_logger
+logger = initial_logger()
+import cv2
+import predict_det
+import predict_rec
+import copy
+import numpy as np
+import math
+import time
+
+
+class TextSystem(object):
+    def __init__(self, args):
+        self.text_detector = predict_det.TextDetector(args)
+        self.text_recognizer = predict_rec.TextRecognizer(args)
+
+    def get_rotate_crop_image(self, img, points):
+        img_height, img_width = img.shape[0:2]
+        left = int(np.min(points[:, 0]))
+        right = int(np.max(points[:, 0]))
+        top = int(np.min(points[:, 1]))
+        bottom = int(np.max(points[:, 1]))
+        img_crop = img[top:bottom, left:right, :].copy()
+        points[:, 0] = points[:, 0] - left
+        points[:, 1] = points[:, 1] - top
+        img_crop_width = int(np.linalg.norm(points[0] - points[1]))
+        img_crop_height = int(np.linalg.norm(points[0] - points[3]))
+        pts_std = np.float32([[0, 0], [img_crop_width, 0],\
+            [img_crop_width, img_crop_height], [0, img_crop_height]])
+        M = cv2.getPerspectiveTransform(points, pts_std)
+        dst_img = cv2.warpPerspective(
+            img_crop,
+            M, (img_crop_width, img_crop_height),
+            borderMode=cv2.BORDER_REPLICATE)
+        dst_img_height, dst_img_width = dst_img.shape[0:2]
+        if dst_img_height * 1.0 / dst_img_width >= 1.5:
+            dst_img = np.rot90(dst_img)
+        return dst_img
+
+    def print_draw_crop_rec_res(self, img_crop_list, rec_res):
+        bbox_num = len(img_crop_list)
+        for bno in range(bbox_num):
+            cv2.imwrite("./output/img_crop_%d.jpg" % bno, img_crop_list[bno])
+            print(bno, rec_res[bno])
+
+    def __call__(self, img):
+        ori_im = img.copy()
+        dt_boxes, elapse = self.text_detector(img)
+        if dt_boxes is None:
+            return None, None
+        img_crop_list = []
+        for bno in range(len(dt_boxes)):
+            tmp_box = copy.deepcopy(dt_boxes[bno])
+            img_crop = self.get_rotate_crop_image(ori_im, tmp_box)
+            img_crop_list.append(img_crop)
+        rec_res, elapse = self.text_recognizer(img_crop_list)
+        #         self.print_draw_crop_rec_res(img_crop_list, rec_res)
+        return dt_boxes, rec_res
+
+
+if __name__ == "__main__":
+    args = utility.parse_args()
+    image_file_list = utility.get_image_file_list(args.image_dir)
+    text_sys = TextSystem(args)
+    for image_file in image_file_list:
+        img = cv2.imread(image_file)
+        if img is None:
+            logger.info("error in loading image:{}".format(image_file))
+            continue
+        starttime = time.time()
+        dt_boxes, rec_res = text_sys(img)
+        elapse = time.time() - starttime
+        print("Predict time of %s: %.3fs" % (image_file, elapse))
+        dt_num = len(dt_boxes)
+        dt_boxes_final = []
+        for dno in range(dt_num):
+            text, score = rec_res[dno]
+            if score >= 0:
+                text_str = "%s, %.3f" % (text, score)
+                print(text_str)
+                dt_boxes_final.append(dt_boxes[dno])
+        utility.draw_text_det_res(dt_boxes_final, image_file)

tools/infer/utility.py

+# Copyright (c) 2020 PaddlePaddle Authors. All Rights Reserved.
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+#     http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+
+import argparse
+import os, sys
+from ppocr.utils.utility import initial_logger
+logger = initial_logger()
+from paddle.fluid.core import PaddleTensor
+from paddle.fluid.core import AnalysisConfig
+from paddle.fluid.core import create_paddle_predictor
+import cv2
+import numpy as np
+
+
+def parse_args():
+    def str2bool(v):
+        return v.lower() in ("true", "t", "1")
+
+    parser = argparse.ArgumentParser()
+    #params for prediction engine
+    parser.add_argument("--use_gpu", type=str2bool, default=True)
+    parser.add_argument("--ir_optim", type=str2bool, default=True)
+    parser.add_argument("--use_tensorrt", type=str2bool, default=False)
+    parser.add_argument("--gpu_mem", type=int, default=8000)
+
+    #params for text detector
+    parser.add_argument("--image_dir", type=str)
+    parser.add_argument("--det_algorithm", type=str, default='DB')
+    parser.add_argument("--det_model_dir", type=str)
+    parser.add_argument("--det_max_side_len", type=float, default=960)
+
+    #DB parmas
+    parser.add_argument("--det_db_thresh", type=float, default=0.3)
+    parser.add_argument("--det_db_box_thresh", type=float, default=0.5)
+
+    #EAST parmas
+    parser.add_argument("--det_east_score_thresh", type=float, default=0.8)
+    parser.add_argument("--det_east_cover_thresh", type=float, default=0.1)
+    parser.add_argument("--det_east_nms_thresh", type=float, default=0.2)
+
+    #params for text recognizer
+    parser.add_argument("--rec_algorithm", type=str, default='CRNN')
+    parser.add_argument("--rec_model_dir", type=str)
+    parser.add_argument("--rec_image_shape", type=str, default="3, 32, 320")
+    parser.add_argument("--rec_char_type", type=str, default='ch')
+    parser.add_argument(
+        "--rec_char_dict_path",
+        type=str,
+        default="./ppocr/utils/ppocr_keys_v1.txt")
+    return parser.parse_args()
+
+
+def get_image_file_list(image_dir):
+    image_file_list = []
+    if image_dir is None:
+        return image_file_list
+    if os.path.isfile(image_dir):
+        image_file_list = [image_dir]
+    elif os.path.isdir(image_dir):
+        for single_file in os.listdir(image_dir):
+            image_file_list.append(os.path.join(image_dir, single_file))
+    return image_file_list
+
+
+def create_predictor(args, mode):
+    if mode == "det":
+        model_dir = args.det_model_dir
+    else:
+        model_dir = args.rec_model_dir
+
+    if model_dir is None:
+        logger.info("not find {} model file path {}".format(mode, model_dir))
+        sys.exit(0)
+    model_file_path = model_dir + "/model"
+    params_file_path = model_dir + "/params"
+    if not os.path.exists(model_file_path):
+        logger.info("not find model file path {}".format(model_file_path))
+        sys.exit(0)
+    if not os.path.exists(params_file_path):
+        logger.info("not find params file path {}".format(params_file_path))
+        sys.exit(0)
+
+    config = AnalysisConfig(model_file_path, params_file_path)
+
+    if args.use_gpu:
+        config.enable_use_gpu(args.gpu_mem, 0)
+    else:
+        config.disable_gpu()
+
+    config.disable_glog_info()
+    config.switch_ir_optim(args.ir_optim)
+    #     if args.use_tensorrt:
+    #         config.enable_tensorrt_engine(
+    #             precision_mode=AnalysisConfig.Precision.Half
+    #             if args.use_fp16 else AnalysisConfig.Precision.Float32,
+    #             max_batch_size=args.batch_size)
+
+    config.enable_memory_optim()
+    # use zero copy
+    config.switch_use_feed_fetch_ops(False)
+    predictor = create_paddle_predictor(config)
+    input_names = predictor.get_input_names()
+    input_tensor = predictor.get_input_tensor(input_names[0])
+    output_names = predictor.get_output_names()
+    output_tensors = []
+    for output_name in output_names:
+        output_tensor = predictor.get_output_tensor(output_name)
+        output_tensors.append(output_tensor)
+    return predictor, input_tensor, output_tensors
+
+
+def draw_text_det_res(dt_boxes, img_path):
+    src_im = cv2.imread(img_path)
+    for box in dt_boxes:
+        box = np.array(box).astype(np.int32).reshape(-1, 2)
+        cv2.polylines(src_im, [box], True, color=(255, 255, 0), thickness=2)
+    img_name_pure = img_path.split("/")[-1]
+    cv2.imwrite("./output/%s" % img_name_pure, src_im)
+
+
+if __name__ == '__main__':
+    args = parse_args()
+    args.use_gpu = False
+    root_path = "/Users/liuweiwei06/Desktop/TEST_CODES/icode/baidu/personal-code/PaddleOCR/"
+    args.det_model_dir = root_path + "test_models/public_v1/ch_det_mv3_db"
+
+    predictor, input_tensor, output_tensors = create_predictor(args, mode='det')
+    print(predictor.get_input_names())
+    print(predictor.get_output_names())
+    print(predictor.program(), file=open("det_program.txt", 'w'))
+
+    args.rec_model_dir = root_path + "test_models/public_v1/ch_rec_mv3_crnn/"
+    rec_predictor, input_tensor, output_tensors = create_predictor(
+        args, mode='rec')
+    print(rec_predictor.get_input_names())
+    print(rec_predictor.get_output_names())

tools/infer_rec.py

+# Copyright (c) 2020 PaddlePaddle Authors. All Rights Reserved.
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+#     http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+
+from __future__ import absolute_import
+from __future__ import division
+from __future__ import print_function
+
+import os
+import time
+import multiprocessing
+import numpy as np
+
+
+def set_paddle_flags(**kwargs):
+    for key, value in kwargs.items():
+        if os.environ.get(key, None) is None:
+            os.environ[key] = str(value)
+
+
+# NOTE(paddle-dev): All of these flags should be
+# set before `import paddle`. Otherwise, it would
+# not take any effect.
+set_paddle_flags(
+    FLAGS_eager_delete_tensor_gb=0,  # enable GC to save memory
+)
+
+from paddle import fluid
+
+# from ppocr.utils.utility import load_config, merge_config
+from ppocr.data.reader_main import test_reader
+import program
+from paddle import fluid
+from ppocr.utils.utility import initial_logger
+logger = initial_logger()
+from ppocr.data.reader_main import reader_main
+from ppocr.utils.save_load import init_model
+from ppocr.utils.character import CharacterOps
+from ppocr.utils.utility import create_module
+
+logger = initial_logger()
+
+
+def main():
+    config = program.load_config(FLAGS.config)
+    program.merge_config(FLAGS.opt)
+    logger.info(config)
+    char_ops = CharacterOps(config['Global'])
+    config['Global']['char_ops'] = char_ops
+
+    # check if set use_gpu=True in paddlepaddle cpu version
+    use_gpu = config['Global']['use_gpu']
+    #     check_gpu(use_gpu)
+
+    place = fluid.CUDAPlace(0) if use_gpu else fluid.CPUPlace()
+    exe = fluid.Executor(place)
+
+    rec_model = create_module(config['Architecture']['function'])(params=config)
+
+    startup_prog = fluid.Program()
+    eval_prog = fluid.Program()
+    with fluid.program_guard(eval_prog, startup_prog):
+        with fluid.unique_name.guard():
+            _, outputs = rec_model(mode="test")
+            fetch_name_list = list(outputs.keys())
+            fetch_varname_list = [outputs[v].name for v in fetch_name_list]
+    eval_prog = eval_prog.clone(for_test=True)
+    exe.run(startup_prog)
+
+    init_model(config, eval_prog, exe)
+
+    blobs = reader_main(config, 'test')
+    imgs = next(blobs())
+    for img in imgs:
+        predict = exe.run(program=eval_prog,
+                          feed={"image": img},
+                          fetch_list=fetch_varname_list,
+                          return_numpy=False)
+
+        preds = np.array(predict[0])
+        if preds.shape[1] == 1:
+            preds = preds.reshape(-1)
+            preds_lod = predict[0].lod()[0]
+            preds_text = char_ops.decode(preds)
+        else:
+            end_pos = np.where(preds[0, :] == 1)[0]
+            if len(end_pos) <= 1:
+                preds_text = preds[0, 1:]
+            else:
+                preds_text = preds[0, 1:end_pos[1]]
+            preds_text = preds_text.reshape(-1)
+            preds_text = char_ops.decode(preds_text)
+
+        print(preds)
+        print(preds_text)
+
+    # save for inference model
+    target_var = []
+    for key, values in outputs.items():
+        target_var.append(values)
+
+    fluid.io.save_inference_model(
+        "./output/",
+        feeded_var_names=['image'],
+        target_vars=target_var,
+        executor=exe,
+        main_program=eval_prog,
+        model_filename="model",
+        params_filename="params")
+
+
+if __name__ == '__main__':
+    parser = program.ArgsParser()
+    FLAGS = parser.parse_args()
+    main()

tools/program.py

+# Copyright (c) 2020 PaddlePaddle Authors. All Rights Reserved.
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+#     http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+
+from __future__ import absolute_import
+from __future__ import division
+from __future__ import print_function
+
+from argparse import ArgumentParser, RawDescriptionHelpFormatter
+import sys
+import yaml
+import os
+from ppocr.utils.utility import create_module
+from ppocr.utils.utility import initial_logger
+logger = initial_logger()
+
+import paddle.fluid as fluid
+import time
+from ppocr.utils.stats import TrainingStats
+from eval_utils.eval_det_utils import eval_det_run
+from eval_utils.eval_rec_utils import eval_rec_run
+from ppocr.utils.save_load import save_model
+import numpy as np
+from ppocr.utils.character import cal_predicts_accuracy
+
+
+class ArgsParser(ArgumentParser):
+    def __init__(self):
+        super(ArgsParser, self).__init__(
+            formatter_class=RawDescriptionHelpFormatter)
+        self.add_argument("-c", "--config", help="configuration file to use")
+        self.add_argument(
+            "-o", "--opt", nargs='+', help="set configuration options")
+
+    def parse_args(self, argv=None):
+        args = super(ArgsParser, self).parse_args(argv)
+        assert args.config is not None, \
+            "Please specify --config=configure_file_path."
+        args.opt = self._parse_opt(args.opt)
+        return args
+
+    def _parse_opt(self, opts):
+        config = {}
+        if not opts:
+            return config
+        for s in opts:
+            s = s.strip()
+            k, v = s.split('=')
+            config[k] = yaml.load(v, Loader=yaml.Loader)
+        return config
+
+
+class AttrDict(dict):
+    """Single level attribute dict, NOT recursive"""
+
+    def __init__(self, **kwargs):
+        super(AttrDict, self).__init__()
+        super(AttrDict, self).update(kwargs)
+
+    def __getattr__(self, key):
+        if key in self:
+            return self[key]
+        raise AttributeError("object has no attribute '{}'".format(key))
+
+
+global_config = AttrDict()
+
+
+def load_config(file_path):
+    """
+    Load config from yml/yaml file.
+
+    Args:
+        file_path (str): Path of the config file to be loaded.
+
+    Returns: global config
+    """
+    _, ext = os.path.splitext(file_path)
+    assert ext in ['.yml', '.yaml'], "only support yaml files for now"
+    merge_config(yaml.load(open(file_path), Loader=yaml.Loader))
+    assert "reader_yml" in global_config['Global'],\
+        "absence reader_yml in global"
+    reader_file_path = global_config['Global']['reader_yml']
+    _, ext = os.path.splitext(reader_file_path)
+    assert ext in ['.yml', '.yaml'], "only support yaml files for reader"
+    merge_config(yaml.load(open(reader_file_path), Loader=yaml.Loader))
+    return global_config
+
+
+def merge_config(config):
+    """
+    Merge config into global config.
+
+    Args:
+        config (dict): Config to be merged.
+
+    Returns: global config
+    """
+    for key, value in config.items():
+        if "." not in key:
+            if isinstance(value, dict) and key in global_config:
+                global_config[key].update(value)
+            else:
+                global_config[key] = value
+        else:
+            sub_keys = key.split('.')
+            assert (sub_keys[0] in global_config)
+            cur = global_config[sub_keys[0]]
+            for idx, sub_key in enumerate(sub_keys[1:]):
+                assert (sub_key in cur)
+                if idx == len(sub_keys) - 2:
+                    cur[sub_key] = value
+                else:
+                    cur = cur[sub_key]
+
+
+def check_gpu(use_gpu):
+    """
+    Log error and exit when set use_gpu=true in paddlepaddle
+    cpu version.
+    """
+    err = "Config use_gpu cannot be set as true while you are " \
+          "using paddlepaddle cpu version ! \nPlease try: \n" \
+          "\t1. Install paddlepaddle-gpu to run model on GPU \n" \
+          "\t2. Set use_gpu as false in config file to run " \
+          "model on CPU"
+
+    try:
+        if use_gpu and not fluid.is_compiled_with_cuda():
+            logger.error(err)
+            sys.exit(1)
+    except Exception as e:
+        pass
+
+
+def build(config, main_prog, startup_prog, mode):
+    """
+    Build a program using a model and an optimizer
+        1. create feeds
+        2. create a dataloader
+        3. create a model
+        4. create fetchs
+        5. create an optimizer
+
+    Args:
+        config(dict): config
+        main_prog(): main program
+        startup_prog(): startup program
+        is_train(bool): train or valid
+
+    Returns:
+        dataloader(): a bridge between the model and the data
+        fetchs(dict): dict of model outputs(included loss and measures)
+    """
+    with fluid.program_guard(main_prog, startup_prog):
+        with fluid.unique_name.guard():
+            func_infor = config['Architecture']['function']
+            model = create_module(func_infor)(params=config)
+            dataloader, outputs = model(mode=mode)
+            fetch_name_list = list(outputs.keys())
+            fetch_varname_list = [outputs[v].name for v in fetch_name_list]
+            opt_loss_name = None
+            if mode == "train":
+                opt_loss = outputs['total_loss']
+                opt_params = config['Optimizer']
+                optimizer = create_module(opt_params['function'])(opt_params)
+                optimizer.minimize(opt_loss)
+                opt_loss_name = opt_loss.name
+                global_lr = optimizer._global_learning_rate()
+                global_lr.persistable = True
+                fetch_name_list.insert(0, "lr")
+                fetch_varname_list.insert(0, global_lr.name)
+    return (dataloader, fetch_name_list, fetch_varname_list, opt_loss_name)
+
+
+def build_export(config, main_prog, startup_prog):
+    """
+    Build a program using a model and an optimizer
+        1. create feeds
+        2. create a dataloader
+        3. create a model
+        4. create fetchs
+        5. create an optimizer
+
+    Args:
+        config(dict): config
+        main_prog(): main program
+        startup_prog(): startup program
+        is_train(bool): train or valid
+
+    Returns:
+        dataloader(): a bridge between the model and the data
+        fetchs(dict): dict of model outputs(included loss and measures)
+    """
+    with fluid.program_guard(main_prog, startup_prog):
+        with fluid.unique_name.guard():
+            func_infor = config['Architecture']['function']
+            model = create_module(func_infor)(params=config)
+            image, outputs = model(mode='export')
+            fetches_var = [outputs[name] for name in outputs]
+            fetches_var_name = [name for name in outputs]
+    feeded_var_names = [image.name]
+    target_vars = fetches_var
+    return feeded_var_names, target_vars, fetches_var_name
+
+
+def create_multi_devices_program(program, loss_var_name):
+    build_strategy = fluid.BuildStrategy()
+    build_strategy.memory_optimize = False
+    build_strategy.enable_inplace = True
+    exec_strategy = fluid.ExecutionStrategy()
+    exec_strategy.num_iteration_per_drop_scope = 1
+    compile_program = fluid.CompiledProgram(program).with_data_parallel(
+        loss_name=loss_var_name,
+        build_strategy=build_strategy,
+        exec_strategy=exec_strategy)
+    return compile_program
+
+
+def train_eval_det_run(config, exe, train_info_dict, eval_info_dict):
+    train_batch_id = 0
+    log_smooth_window = config['Global']['log_smooth_window']
+    epoch_num = config['Global']['epoch_num']
+    print_batch_step = config['Global']['print_batch_step']
+    eval_batch_step = config['Global']['eval_batch_step']
+    save_epoch_step = config['Global']['save_epoch_step']
+    save_model_dir = config['Global']['save_model_dir']
+    train_stats = TrainingStats(log_smooth_window,
+                                train_info_dict['fetch_name_list'])
+    best_eval_hmean = -1
+    best_batch_id = 0
+    best_epoch = 0
+    train_loader = train_info_dict['reader']
+    for epoch in range(epoch_num):
+        train_loader.start()
+        try:
+            while True:
+                t1 = time.time()
+                train_outs = exe.run(
+                    program=train_info_dict['compile_program'],
+                    fetch_list=train_info_dict['fetch_varname_list'],
+                    return_numpy=False)
+                stats = {}
+                for tno in range(len(train_outs)):
+                    fetch_name = train_info_dict['fetch_name_list'][tno]
+                    fetch_value = np.mean(np.array(train_outs[tno]))
+                    stats[fetch_name] = fetch_value
+                t2 = time.time()
+                train_batch_elapse = t2 - t1
+                train_stats.update(stats)
+                if train_batch_id > 0 and train_batch_id \
+                    % print_batch_step == 0:
+                    logs = train_stats.log()
+                    strs = 'epoch: {}, iter: {}, {}, time: {:.3f}'.format(
+                        epoch, train_batch_id, logs, train_batch_elapse)
+                    logger.info(strs)
+
+                if train_batch_id > 0 and\
+                    train_batch_id % eval_batch_step == 0:
+                    metrics = eval_det_run(exe, config, eval_info_dict, "eval")
+                    hmean = metrics['hmean']
+                    if hmean >= best_eval_hmean:
+                        best_eval_hmean = hmean
+                        best_batch_id = train_batch_id
+                        best_epoch = epoch
+                        save_path = save_model_dir + "/best_accuracy"
+                        save_model(train_info_dict['train_program'], save_path)
+                    strs = 'Test iter: {}, metrics:{}, best_hmean:{:.6f}, best_epoch:{}, best_batch_id:{}'.format(
+                        train_batch_id, metrics, best_eval_hmean, best_epoch,
+                        best_batch_id)
+                    logger.info(strs)
+                train_batch_id += 1
+
+        except fluid.core.EOFException:
+            train_loader.reset()
+
+        if epoch > 0 and epoch % save_epoch_step == 0:
+            save_path = save_model_dir + "/iter_epoch_%d" % (epoch)
+            save_model(train_info_dict['train_program'], save_path)
+    return
+
+
+def train_eval_rec_run(config, exe, train_info_dict, eval_info_dict):
+    train_batch_id = 0
+    log_smooth_window = config['Global']['log_smooth_window']
+    epoch_num = config['Global']['epoch_num']
+    print_batch_step = config['Global']['print_batch_step']
+    eval_batch_step = config['Global']['eval_batch_step']
+    save_epoch_step = config['Global']['save_epoch_step']
+    save_model_dir = config['Global']['save_model_dir']
+    train_stats = TrainingStats(log_smooth_window, ['loss', 'acc'])
+    best_eval_acc = -1
+    best_batch_id = 0
+    best_epoch = 0
+    train_loader = train_info_dict['reader']
+    for epoch in range(epoch_num):
+        train_loader.start()
+        try:
+            while True:
+                t1 = time.time()
+                train_outs = exe.run(
+                    program=train_info_dict['compile_program'],
+                    fetch_list=train_info_dict['fetch_varname_list'],
+                    return_numpy=False)
+                fetch_map = dict(
+                    zip(train_info_dict['fetch_name_list'],
+                        range(len(train_outs))))
+
+                loss = np.mean(np.array(train_outs[fetch_map['total_loss']]))
+                lr = np.mean(np.array(train_outs[fetch_map['lr']]))
+                preds_idx = fetch_map['decoded_out']
+                preds = np.array(train_outs[preds_idx])
+                preds_lod = train_outs[preds_idx].lod()[0]
+                labels_idx = fetch_map['label']
+                labels = np.array(train_outs[labels_idx])
+                labels_lod = train_outs[labels_idx].lod()[0]
+
+                acc, acc_num, img_num = cal_predicts_accuracy(
+                    config['Global']['char_ops'], preds, preds_lod, labels,
+                    labels_lod)
+                t2 = time.time()
+                train_batch_elapse = t2 - t1
+                stats = {'loss': loss, 'acc': acc}
+                train_stats.update(stats)
+                if train_batch_id > 0 and train_batch_id \
+                    % print_batch_step == 0:
+                    logs = train_stats.log()
+                    strs = 'epoch: {}, iter: {}, lr: {:.6f}, {}, time: {:.3f}'.format(
+                        epoch, train_batch_id, lr, logs, train_batch_elapse)
+                    logger.info(strs)
+
+                if train_batch_id > 0 and\
+                    train_batch_id % eval_batch_step == 0:
+                    metrics = eval_rec_run(exe, config, eval_info_dict, "eval")
+                    eval_acc = metrics['avg_acc']
+                    eval_sample_num = metrics['total_sample_num']
+                    if eval_acc > best_eval_acc:
+                        best_eval_acc = eval_acc
+                        best_batch_id = train_batch_id
+                        best_epoch = epoch
+                        save_path = save_model_dir + "/best_accuracy"
+                        save_model(train_info_dict['train_program'], save_path)
+                    strs = 'Test iter: {}, acc:{:.6f}, best_acc:{:.6f}, best_epoch:{}, best_batch_id:{}, eval_sample_num:{}'.format(
+                        train_batch_id, eval_acc, best_eval_acc, best_epoch,
+                        best_batch_id, eval_sample_num)
+                    logger.info(strs)
+                train_batch_id += 1
+
+        except fluid.core.EOFException:
+            train_loader.reset()
+
+        if epoch > 0 and epoch % save_epoch_step == 0:
+            save_path = save_model_dir + "/iter_epoch_%d" % (epoch)
+            save_model(train_info_dict['train_program'], save_path)
+    return

tools/tmp/eval_det.py

+# Copyright (c) 2020 PaddlePaddle Authors. All Rights Reserved.
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+#     http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+
+from __future__ import absolute_import
+from __future__ import division
+from __future__ import print_function
+
+import os
+import sys
+import time
+import numpy as np
+from copy import deepcopy
+import json
+
+# from paddle.fluid.contrib.model_stat import summary
+
+
+def set_paddle_flags(**kwargs):
+    for key, value in kwargs.items():
+        if os.environ.get(key, None) is None:
+            os.environ[key] = str(value)
+
+
+# NOTE(paddle-dev): All of these flags should be
+# set before `import paddle`. Otherwise, it would
+# not take any effect. 
+set_paddle_flags(
+    FLAGS_eager_delete_tensor_gb=0,  # enable GC to save memory
+)
+
+from paddle import fluid
+from ppocr.utils.utility import create_module
+from ppocr.utils.utility import load_config, merge_config
+import ppocr.data.det.reader_main as reader
+from ppocr.utils.utility import ArgsParser
+from ppocr.utils.check import check_gpu
+from ppocr.utils.checkpoint import load_pretrain, load_checkpoint, save, save_model
+
+from ppocr.utils.utility import initial_logger
+logger = initial_logger()
+from ppocr.utils.eval_utils import eval_det_run
+
+
+def draw_det_res(dt_boxes, config, img_name, ino):
+    if len(dt_boxes) > 0:
+        img_set_path = config['TestReader']['img_set_dir']
+        img_path = img_set_path + img_name
+        import cv2
+        src_im = cv2.imread(img_path)
+        for box in dt_boxes:
+            box = box.astype(np.int32).reshape((-1, 1, 2))
+            cv2.polylines(src_im, [box], True, color=(255, 255, 0), thickness=2)
+        cv2.imwrite("tmp%d.jpg" % ino, src_im)
+
+
+def main():
+    config = load_config(FLAGS.config)
+    merge_config(FLAGS.opt)
+    print(config)
+
+    # check if set use_gpu=True in paddlepaddle cpu version
+    use_gpu = config['Global']['use_gpu']
+    check_gpu(use_gpu)
+
+    place = fluid.CUDAPlace(0) if use_gpu else fluid.CPUPlace()
+    exe = fluid.Executor(place)
+
+    det_model = create_module(config['Architecture']['function'])(params=config)
+
+    startup_prog = fluid.Program()
+    eval_prog = fluid.Program()
+    with fluid.program_guard(eval_prog, startup_prog):
+        with fluid.unique_name.guard():
+            eval_loader, eval_outputs = det_model(mode="test")
+            eval_fetch_list = [v.name for v in eval_outputs]
+    eval_prog = eval_prog.clone(for_test=True)
+    exe.run(startup_prog)
+
+    pretrain_weights = config['Global']['pretrain_weights']
+    if pretrain_weights is not None:
+        load_pretrain(exe, eval_prog, pretrain_weights)
+#         fluid.load(eval_prog, pretrain_weights)
+#         def if_exist(var):
+#             return os.path.exists(os.path.join(pretrain_weights, var.name))
+#         fluid.io.load_vars(exe, pretrain_weights, predicate=if_exist, main_program=eval_prog)
+    else:
+        logger.info("Not find pretrain_weights:%s" % pretrain_weights)
+        sys.exit(0)
+
+#     fluid.io.save_inference_model("./output/", feeded_var_names=['image'],
+#         target_vars=eval_outputs, executor=exe, main_program=eval_prog,
+#         model_filename="model", params_filename="params")
+#     sys.exit(-1)
+
+    metrics = eval_det_run(exe, eval_prog, eval_fetch_list, config, "test")
+    logger.info("metrics:{}".format(metrics))
+    logger.info("success!")
+
+
+def test_reader():
+    config = load_config(FLAGS.config)
+    merge_config(FLAGS.opt)
+    print(config)
+    tmp_reader = reader.test_reader(config=config)
+    count = 0
+    print_count = 0
+    import time
+    starttime = time.time()
+    for data in tmp_reader():
+        count += len(data)
+        print_count += 1
+        if print_count % 10 == 0:
+            batch_time = (time.time() - starttime) / print_count
+            print("reader:", count, len(data), batch_time)
+    print("finish reader:", count)
+    print("success")
+
+
+if __name__ == '__main__':
+    parser = ArgsParser()
+    FLAGS = parser.parse_args()
+    main()
+#     test_reader()

tools/tmp/infer_det.py

+# Copyright (c) 2020 PaddlePaddle Authors. All Rights Reserved.
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+#     http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+
+from __future__ import absolute_import
+from __future__ import division
+from __future__ import print_function
+
+import os
+import sys
+import time
+import numpy as np
+from copy import deepcopy
+import json
+
+# from paddle.fluid.contrib.model_stat import summary
+
+
+def set_paddle_flags(**kwargs):
+    for key, value in kwargs.items():
+        if os.environ.get(key, None) is None:
+            os.environ[key] = str(value)
+
+
+# NOTE(paddle-dev): All of these flags should be
+# set before `import paddle`. Otherwise, it would
+# not take any effect. 
+set_paddle_flags(
+    FLAGS_eager_delete_tensor_gb=0,  # enable GC to save memory
+)
+
+from paddle import fluid
+from ppocr.utils.utility import create_module
+from ppocr.utils.utility import load_config, merge_config
+import ppocr.data.det.reader_main as reader
+from ppocr.utils.utility import ArgsParser
+from ppocr.utils.check import check_gpu
+from ppocr.utils.checkpoint import load_pretrain, load_checkpoint, save, save_model
+
+from ppocr.utils.utility import initial_logger
+logger = initial_logger()
+from ppocr.utils.eval_utils import eval_det_run
+
+
+def draw_det_res(dt_boxes, config, img_name, ino):
+    if len(dt_boxes) > 0:
+        img_set_path = config['TestReader']['img_set_dir']
+        img_path = img_set_path + img_name
+        import cv2
+        src_im = cv2.imread(img_path)
+        for box in dt_boxes:
+            box = box.astype(np.int32).reshape((-1, 1, 2))
+            cv2.polylines(src_im, [box], True, color=(255, 255, 0), thickness=2)
+        cv2.imwrite("tmp%d.jpg" % ino, src_im)
+
+
+def main():
+    config = load_config(FLAGS.config)
+    merge_config(FLAGS.opt)
+    print(config)
+
+    # check if set use_gpu=True in paddlepaddle cpu version
+    use_gpu = config['Global']['use_gpu']
+    check_gpu(use_gpu)
+
+    place = fluid.CUDAPlace(0) if use_gpu else fluid.CPUPlace()
+    exe = fluid.Executor(place)
+
+    det_model = create_module(config['Architecture']['function'])(params=config)
+
+    startup_prog = fluid.Program()
+    eval_prog = fluid.Program()
+    with fluid.program_guard(eval_prog, startup_prog):
+        with fluid.unique_name.guard():
+            eval_outputs = det_model(mode="test")
+            eval_fetch_list = [v.name for v in eval_outputs]
+    eval_prog = eval_prog.clone(for_test=True)
+    exe.run(startup_prog)
+
+    pretrain_weights = config['Global']['pretrain_weights']
+    if pretrain_weights is not None:
+        fluid.load(eval_prog, pretrain_weights)
+    else:
+        logger.info("Not find pretrain_weights:%s" % pretrain_weights)
+        sys.exit(0)
+
+    save_res_path = config['Global']['save_res_path']
+    with open(save_res_path, "wb") as fout:
+        test_reader = reader.test_reader(config=config)
+        tackling_num = 0
+        for data in test_reader():
+            img_num = len(data)
+            tackling_num = tackling_num + img_num
+            logger.info("tackling_num:%d", tackling_num)
+            img_list = []
+            ratio_list = []
+            img_name_list = []
+            for ino in range(img_num):
+                img_list.append(data[ino][0])
+                ratio_list.append(data[ino][1])
+                img_name_list.append(data[ino][2])
+            img_list = np.concatenate(img_list, axis=0)
+            outs = exe.run(eval_prog,\
+                feed={'image': img_list},\
+                fetch_list=eval_fetch_list)
+
+            global_params = config['Global']
+            postprocess_params = deepcopy(config["PostProcess"])
+            postprocess_params.update(global_params)
+            postprocess = create_module(postprocess_params['function'])\
+                (params=postprocess_params)
+            dt_boxes_list = postprocess(outs, ratio_list)
+            for ino in range(img_num):
+                dt_boxes = dt_boxes_list[ino]
+                img_name = img_name_list[ino]
+                dt_boxes_json = []
+                for box in dt_boxes:
+                    tmp_json = {"transcription": ""}
+                    tmp_json['points'] = box.tolist()
+                    dt_boxes_json.append(tmp_json)
+                otstr = img_name + "\t" + json.dumps(dt_boxes_json) + "\n"
+                fout.write(otstr.encode())
+                #draw_det_res(dt_boxes, config, img_name, ino)
+    logger.info("success!")
+
+
+def test_reader():
+    config = load_config(FLAGS.config)
+    merge_config(FLAGS.opt)
+    print(config)
+    tmp_reader = reader.test_reader(config=config)
+    count = 0
+    print_count = 0
+    import time
+    starttime = time.time()
+    for data in tmp_reader():
+        count += len(data)
+        print_count += 1
+        if print_count % 10 == 0:
+            batch_time = (time.time() - starttime) / print_count
+            print("reader:", count, len(data), batch_time)
+    print("finish reader:", count)
+    print("success")
+
+
+if __name__ == '__main__':
+    parser = ArgsParser()
+    FLAGS = parser.parse_args()
+    main()
+#     test_reader()

tools/tmp/infer_rec.py

+# Copyright (c) 2020 PaddlePaddle Authors. All Rights Reserved.
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+#     http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+
+from __future__ import absolute_import
+from __future__ import division
+from __future__ import print_function
+
+import os
+import time
+import multiprocessing
+import numpy as np
+
+
+def set_paddle_flags(**kwargs):
+    for key, value in kwargs.items():
+        if os.environ.get(key, None) is None:
+            os.environ[key] = str(value)
+
+
+# NOTE(paddle-dev): All of these flags should be
+# set before `import paddle`. Otherwise, it would
+# not take any effect. 
+set_paddle_flags(
+    FLAGS_eager_delete_tensor_gb=0,  # enable GC to save memory
+)
+
+from paddle import fluid
+
+from ppocr.utils.utility import load_config, merge_config
+from ppocr.data.rec.reader_main import test_reader
+
+from ppocr.utils.utility import ArgsParser
+from ppocr.utils.character import CharacterOps, cal_predicts_accuracy
+from ppocr.utils.check import check_gpu
+from ppocr.utils.utility import create_module
+
+from ppocr.utils.utility import initial_logger
+logger = initial_logger()
+
+
+def main():
+    config = load_config(FLAGS.config)
+    merge_config(FLAGS.opt)
+    char_ops = CharacterOps(config['Global'])
+    config['Global']['char_num'] = char_ops.get_char_num()
+
+    # check if set use_gpu=True in paddlepaddle cpu version
+    use_gpu = config['Global']['use_gpu']
+    check_gpu(use_gpu)
+
+    place = fluid.CUDAPlace(0) if use_gpu else fluid.CPUPlace()
+    exe = fluid.Executor(place)
+
+    rec_model = create_module(config['Architecture']['function'])(params=config)
+
+    startup_prog = fluid.Program()
+    eval_prog = fluid.Program()
+    with fluid.program_guard(eval_prog, startup_prog):
+        with fluid.unique_name.guard():
+            eval_outputs = rec_model(mode="test")
+            eval_fetch_list = [v.name for v in eval_outputs]
+    eval_prog = eval_prog.clone(for_test=True)
+    exe.run(startup_prog)
+
+    pretrain_weights = config['Global']['pretrain_weights']
+    if pretrain_weights is not None:
+        fluid.load(eval_prog, pretrain_weights)
+
+    test_img_path = config['test_img_path']
+    image_shape = config['Global']['image_shape']
+    blobs = test_reader(image_shape, test_img_path)
+    predict = exe.run(program=eval_prog,
+                      feed={"image": blobs},
+                      fetch_list=eval_fetch_list,
+                      return_numpy=False)
+    preds = np.array(predict[0])
+    if preds.shape[1] == 1:
+        preds = preds.reshape(-1)
+        preds_lod = predict[0].lod()[0]
+        preds_text = char_ops.decode(preds)
+    else:
+        end_pos = np.where(preds[0, :] == 1)[0]
+        if len(end_pos) <= 1:
+            preds_text = preds[0, 1:]
+        else:
+            preds_text = preds[0, 1:end_pos[1]]
+        preds_text = preds_text.reshape(-1)
+        preds_text = char_ops.decode(preds_text)
+
+    fluid.io.save_inference_model(
+        "./output/",
+        feeded_var_names=['image'],
+        target_vars=eval_outputs,
+        executor=exe,
+        main_program=eval_prog,
+        model_filename="model",
+        params_filename="params")
+    print(preds)
+    print(preds_text)
+
+
+if __name__ == '__main__':
+    parser = ArgsParser()
+    FLAGS = parser.parse_args()
+    main()