Skip to content

GitLab

Commit 78d51971 authored by Leif's avatar Leif
Browse files

Merge remote-tracking branch 'upstream/dygraph' into dy3

parents bd314018 c683a181
Hide whitespace changes
Inline Side-by-side
Showing with 112 additions and 736 deletions
deploy/pdserving/det_web_server.py deleted 100644 → 0
View file @ bd314018
# 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 paddle_serving_client import Client
import cv2
import sys
import numpy as np
import os
from paddle_serving_client import Client
from paddle_serving_app.reader import Sequential, ResizeByFactor
from paddle_serving_app.reader import Div, Normalize, Transpose
from paddle_serving_app.reader import DBPostProcess, FilterBoxes
if sys.argv[1] == 'gpu':
from paddle_serving_server_gpu.web_service import WebService
elif sys.argv[1] == 'cpu':
from paddle_serving_server.web_service import WebService
import time
import re
import base64
class OCRService(WebService):
def init_det(self):
self.det_preprocess = Sequential([
ResizeByFactor(32, 960), Div(255),
Normalize([0.485, 0.456, 0.406], [0.229, 0.224, 0.225]), Transpose(
(2, 0, 1))
])
self.filter_func = FilterBoxes(10, 10)
self.post_func = DBPostProcess({
"thresh": 0.3,
"box_thresh": 0.5,
"max_candidates": 1000,
"unclip_ratio": 1.5,
"min_size": 3
})
def preprocess(self, feed=[], fetch=[]):
data = base64.b64decode(feed[0]["image"].encode('utf8'))
data = np.fromstring(data, np.uint8)
im = cv2.imdecode(data, cv2.IMREAD_COLOR)
self.ori_h, self.ori_w, _ = im.shape
det_img = self.det_preprocess(im)
_, self.new_h, self.new_w = det_img.shape
print(det_img)
return {"image": det_img}, ["concat_1.tmp_0"]
def postprocess(self, feed={}, fetch=[], fetch_map=None):
det_out = fetch_map["concat_1.tmp_0"]
ratio_list = [
float(self.new_h) / self.ori_h, float(self.new_w) / self.ori_w
]
dt_boxes_list = self.post_func(det_out, [ratio_list])
dt_boxes = self.filter_func(dt_boxes_list[0], [self.ori_h, self.ori_w])
return {"dt_boxes": dt_boxes.tolist()}
ocr_service = OCRService(name="ocr")
ocr_service.load_model_config("ocr_det_model")
if sys.argv[1] == 'gpu':
ocr_service.set_gpus("0")
ocr_service.prepare_server(workdir="workdir", port=9292, device="gpu", gpuid=0)
elif sys.argv[1] == 'cpu':
ocr_service.prepare_server(workdir="workdir", port=9292, device="cpu")
ocr_service.init_det()
ocr_service.run_rpc_service()
ocr_service.run_web_service()
deploy/pdserving/ocr_local_server.py deleted 100644 → 0
View file @ bd314018
# 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 paddle_serving_client import Client
from paddle_serving_app.reader import OCRReader
import cv2
import sys
import numpy as np
import os
from paddle_serving_client import Client
from paddle_serving_app.reader import Sequential, URL2Image, ResizeByFactor
from paddle_serving_app.reader import Div, Normalize, Transpose
from paddle_serving_app.reader import DBPostProcess, FilterBoxes, GetRotateCropImage, SortedBoxes
if sys.argv[1] == 'gpu':
from paddle_serving_server_gpu.web_service import WebService
elif sys.argv[1] == 'cpu':
from paddle_serving_server.web_service import WebService
from paddle_serving_app.local_predict import Debugger
import time
import re
import base64
class OCRService(WebService):
def init_det_debugger(self, det_model_config):
self.det_preprocess = Sequential([
ResizeByFactor(32, 960), Div(255),
Normalize([0.485, 0.456, 0.406], [0.229, 0.224, 0.225]), Transpose(
(2, 0, 1))
])
self.det_client = Debugger()
if sys.argv[1] == 'gpu':
self.det_client.load_model_config(
det_model_config, gpu=True, profile=False)
elif sys.argv[1] == 'cpu':
self.det_client.load_model_config(
det_model_config, gpu=False, profile=False)
self.ocr_reader = OCRReader()
def preprocess(self, feed=[], fetch=[]):
data = base64.b64decode(feed[0]["image"].encode('utf8'))
data = np.fromstring(data, np.uint8)
im = cv2.imdecode(data, cv2.IMREAD_COLOR)
ori_h, ori_w, _ = im.shape
det_img = self.det_preprocess(im)
_, new_h, new_w = det_img.shape
det_img = det_img[np.newaxis, :]
det_img = det_img.copy()
det_out = self.det_client.predict(
feed={"image": det_img}, fetch=["concat_1.tmp_0"])
filter_func = FilterBoxes(10, 10)
post_func = DBPostProcess({
"thresh": 0.3,
"box_thresh": 0.5,
"max_candidates": 1000,
"unclip_ratio": 1.5,
"min_size": 3
})
sorted_boxes = SortedBoxes()
ratio_list = [float(new_h) / ori_h, float(new_w) / ori_w]
dt_boxes_list = post_func(det_out["concat_1.tmp_0"], [ratio_list])
dt_boxes = filter_func(dt_boxes_list[0], [ori_h, ori_w])
dt_boxes = sorted_boxes(dt_boxes)
get_rotate_crop_image = GetRotateCropImage()
img_list = []
max_wh_ratio = 0
for i, dtbox in enumerate(dt_boxes):
boximg = get_rotate_crop_image(im, dt_boxes[i])
img_list.append(boximg)
h, w = boximg.shape[0:2]
wh_ratio = w * 1.0 / h
max_wh_ratio = max(max_wh_ratio, wh_ratio)
if len(img_list) == 0:
return [], []
_, w, h = self.ocr_reader.resize_norm_img(img_list[0],
max_wh_ratio).shape
imgs = np.zeros((len(img_list), 3, w, h)).astype('float32')
for id, img in enumerate(img_list):
norm_img = self.ocr_reader.resize_norm_img(img, max_wh_ratio)
imgs[id] = norm_img
feed = {"image": imgs.copy()}
fetch = ["ctc_greedy_decoder_0.tmp_0", "softmax_0.tmp_0"]
return feed, fetch
def postprocess(self, feed={}, fetch=[], fetch_map=None):
rec_res = self.ocr_reader.postprocess(fetch_map, with_score=True)
res_lst = []
for res in rec_res:
res_lst.append(res[0])
res = {"res": res_lst}
return res
ocr_service = OCRService(name="ocr")
ocr_service.load_model_config("ocr_rec_model")
ocr_service.init_det_debugger(det_model_config="ocr_det_model")
if sys.argv[1] == 'gpu':
ocr_service.prepare_server(workdir="workdir", port=9292, device="gpu", gpuid=0)
ocr_service.run_debugger_service(gpu=True)
elif sys.argv[1] == 'cpu':
ocr_service.prepare_server(workdir="workdir", port=9292, device="cpu")
ocr_service.run_debugger_service()
ocr_service.run_web_service()
deploy/pdserving/ocr_web_server.py deleted 100644 → 0
View file @ bd314018
# 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 paddle_serving_client import Client
from paddle_serving_app.reader import OCRReader
import cv2
import sys
import numpy as np
import os
from paddle_serving_client import Client
from paddle_serving_app.reader import Sequential, URL2Image, ResizeByFactor
from paddle_serving_app.reader import Div, Normalize, Transpose
from paddle_serving_app.reader import DBPostProcess, FilterBoxes, GetRotateCropImage, SortedBoxes
if sys.argv[1] == 'gpu':
from paddle_serving_server_gpu.web_service import WebService
elif sys.argv[1] == 'cpu':
from paddle_serving_server.web_service import WebService
import time
import re
import base64
class OCRService(WebService):
def init_det_client(self, det_port, det_client_config):
self.det_preprocess = Sequential([
ResizeByFactor(32, 960), Div(255),
Normalize([0.485, 0.456, 0.406], [0.229, 0.224, 0.225]), Transpose(
(2, 0, 1))
])
self.det_client = Client()
self.det_client.load_client_config(det_client_config)
self.det_client.connect(["127.0.0.1:{}".format(det_port)])
self.ocr_reader = OCRReader()
def preprocess(self, feed=[], fetch=[]):
data = base64.b64decode(feed[0]["image"].encode('utf8'))
data = np.fromstring(data, np.uint8)
im = cv2.imdecode(data, cv2.IMREAD_COLOR)
ori_h, ori_w, _ = im.shape
det_img = self.det_preprocess(im)
det_out = self.det_client.predict(
feed={"image": det_img}, fetch=["concat_1.tmp_0"])
_, new_h, new_w = det_img.shape
filter_func = FilterBoxes(10, 10)
post_func = DBPostProcess({
"thresh": 0.3,
"box_thresh": 0.5,
"max_candidates": 1000,
"unclip_ratio": 1.5,
"min_size": 3
})
sorted_boxes = SortedBoxes()
ratio_list = [float(new_h) / ori_h, float(new_w) / ori_w]
dt_boxes_list = post_func(det_out["concat_1.tmp_0"], [ratio_list])
dt_boxes = filter_func(dt_boxes_list[0], [ori_h, ori_w])
dt_boxes = sorted_boxes(dt_boxes)
get_rotate_crop_image = GetRotateCropImage()
feed_list = []
img_list = []
max_wh_ratio = 0
for i, dtbox in enumerate(dt_boxes):
boximg = get_rotate_crop_image(im, dt_boxes[i])
img_list.append(boximg)
h, w = boximg.shape[0:2]
wh_ratio = w * 1.0 / h
max_wh_ratio = max(max_wh_ratio, wh_ratio)
for img in img_list:
norm_img = self.ocr_reader.resize_norm_img(img, max_wh_ratio)
feed = {"image": norm_img}
feed_list.append(feed)
fetch = ["ctc_greedy_decoder_0.tmp_0", "softmax_0.tmp_0"]
return feed_list, fetch
def postprocess(self, feed={}, fetch=[], fetch_map=None):
rec_res = self.ocr_reader.postprocess(fetch_map, with_score=True)
res_lst = []
for res in rec_res:
res_lst.append(res[0])
res = {"res": res_lst}
return res
ocr_service = OCRService(name="ocr")
ocr_service.load_model_config("ocr_rec_model")
if sys.argv[1] == 'gpu':
ocr_service.set_gpus("0")
ocr_service.prepare_server(workdir="workdir", port=9292, device="gpu", gpuid=0)
elif sys.argv[1] == 'cpu':
ocr_service.prepare_server(workdir="workdir", port=9292)
ocr_service.init_det_client(
det_port=9293,
det_client_config="ocr_det_client/serving_client_conf.prototxt")
ocr_service.run_rpc_service()
ocr_service.run_web_service()
deploy/pdserving/readme.md deleted 100644 → 0
View file @ bd314018
# Paddle Serving 服务部署(Beta)
本教程将介绍基于[Paddle Serving](https://github.com/PaddlePaddle/Serving)部署PaddleOCR在线预测服务的详细步骤。
## 快速启动服务
### 1. 准备环境
我们先安装Paddle Serving相关组件
我们推荐用户使用GPU来做Paddle Serving的OCR服务部署
**CUDA版本：9.0**
**CUDNN版本：7.0**
**操作系统版本：CentOS 6以上**
**Python3操作指南：**
```
#以下提供beta版本的paddle serving whl包，欢迎试用，正式版会在8月中正式上线
#GPU用户下载server包使用这个链接
wget --no-check-certificate https://paddle-serving.bj.bcebos.com/others/paddle_serving_server_gpu-0.3.2-py3-none-any.whl
python -m pip install paddle_serving_server_gpu-0.3.2-py3-none-any.whl
#CPU版本使用这个链接
wget --no-check-certificate https://paddle-serving.bj.bcebos.com/others/paddle_serving_server-0.3.2-py3-none-any.whl
python -m pip install paddle_serving_server-0.3.2-py3-none-any.whl
#客户端和App包使用以下链接（CPU，GPU通用）
wget --no-check-certificate https://paddle-serving.bj.bcebos.com/others/paddle_serving_client-0.3.2-cp36-none-any.whl
wget --no-check-certificate https://paddle-serving.bj.bcebos.com/others/paddle_serving_app-0.1.2-py3-none-any.whl
python -m pip install paddle_serving_app-0.1.2-py3-none-any.whl paddle_serving_client-0.3.2-cp36-none-any.whl
```
**Python2操作指南：**
```
#以下提供beta版本的paddle serving whl包，欢迎试用，正式版会在8月中正式上线
#GPU用户下载server包使用这个链接
wget --no-check-certificate https://paddle-serving.bj.bcebos.com/others/paddle_serving_server_gpu-0.3.2-py2-none-any.whl
python -m pip install paddle_serving_server_gpu-0.3.2-py2-none-any.whl
#CPU版本使用这个链接
wget --no-check-certificate https://paddle-serving.bj.bcebos.com/others/paddle_serving_server-0.3.2-py2-none-any.whl
python -m pip install paddle_serving_server-0.3.2-py2-none-any.whl
#客户端和App包使用以下链接（CPU，GPU通用）
wget --no-check-certificate https://paddle-serving.bj.bcebos.com/others/paddle_serving_app-0.1.2-py2-none-any.whl
wget --no-check-certificate https://paddle-serving.bj.bcebos.com/others/paddle_serving_client-0.3.2-cp27-none-any.whl
python -m pip install paddle_serving_app-0.1.2-py2-none-any.whl paddle_serving_client-0.3.2-cp27-none-any.whl
```
### 2. 模型转换
可以使用`paddle_serving_app`提供的模型，执行下列命令
```
python -m paddle_serving_app.package --get_model ocr_rec
tar -xzvf ocr_rec.tar.gz
python -m paddle_serving_app.package --get_model ocr_det
tar -xzvf ocr_det.tar.gz
```
执行上述命令会下载`db_crnn_mobile`的模型，如果想要下载规模更大的`db_crnn_server`模型，可以在下载预测模型并解压之后。参考[如何从Paddle保存的预测模型转为Paddle Serving格式可部署的模型](https://github.com/PaddlePaddle/Serving/blob/develop/doc/INFERENCE_TO_SERVING_CN.md)。
我们以`ch_rec_r34_vd_crnn`模型作为例子，下载链接在：
```
wget --no-check-certificate https://paddleocr.bj.bcebos.com/ch_models/ch_rec_r34_vd_crnn_infer.tar
tar xf ch_rec_r34_vd_crnn_infer.tar
```
因此我们按照Serving模型转换教程，运行下列python文件。
```
from paddle_serving_client.io import inference_model_to_serving
inference_model_dir = "ch_rec_r34_vd_crnn"
serving_client_dir = "serving_client_dir"
serving_server_dir = "serving_server_dir"
feed_var_names, fetch_var_names = inference_model_to_serving(
inference_model_dir, serving_client_dir, serving_server_dir, model_filename="model", params_filename="params")
```
最终会在`serving_client_dir`和`serving_server_dir`生成客户端和服务端的模型配置。
### 3. 启动服务
启动服务可以根据实际需求选择启动`标准版`或者`快速版`，两种方式的对比如下表：
|版本|特点|适用场景|
|-|-|-|
|标准版|稳定性高，分布式部署|适用于吞吐量大，需要跨机房部署的情况|
|快速版|部署方便，预测速度快|适用于对预测速度要求高，迭代速度快的场景|
#### 方式1. 启动标准版服务
```
# cpu，gpu启动二选一，以下是cpu启动
python -m paddle_serving_server.serve --model ocr_det_model --port 9293
python ocr_web_server.py cpu
# gpu启动
python -m paddle_serving_server_gpu.serve --model ocr_det_model --port 9293 --gpu_id 0
python ocr_web_server.py gpu
```
#### 方式2. 启动快速版服务
```
# cpu，gpu启动二选一，以下是cpu启动
python ocr_local_server.py cpu
# gpu启动
python ocr_local_server.py gpu
```
## 发送预测请求
```
python ocr_web_client.py
```
## 返回结果格式说明
返回结果是json格式
```
{u'result': {u'res': [u'\u571f\u5730\u6574\u6cbb\u4e0e\u571f\u58e4\u4fee\u590d\u7814\u7a76\u4e2d\u5fc3', u'\u534e\u5357\u519c\u4e1a\u5927\u5b661\u7d20\u56fe']}}
```
我们也可以打印结果json串中`res`字段的每一句话
```
土地整治与土壤修复研究中心
华南农业大学1素图
```
## 自定义修改服务逻辑
在`ocr_web_server.py`或是`ocr_local_server.py`当中的`preprocess`函数里面做了检测服务和识别服务的前处理，`postprocess`函数里面做了识别的后处理服务，可以在相应的函数中做修改。调用了`paddle_serving_app`库提供的常见CV模型的前处理/后处理库。
如果想要单独启动Paddle Serving的检测服务和识别服务，参见下列表格, 执行对应的脚本即可，并且在命令行参数注明用的CPU或是GPU来提供服务。
| 模型 | 标准版 | 快速版 |
| ---- | ----------------- | ------------------- |
| 检测 | det_web_server.py | det_local_server.py |
| 识别 | rec_web_server.py | rec_local_server.py |
更多信息参见[Paddle Serving](https://github.com/PaddlePaddle/Serving)
deploy/pdserving/rec_local_server.py deleted 100644 → 0
View file @ bd314018
# 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 paddle_serving_client import Client
from paddle_serving_app.reader import OCRReader
import cv2
import sys
import numpy as np
import os
from paddle_serving_client import Client
from paddle_serving_app.reader import Sequential, URL2Image, ResizeByFactor
from paddle_serving_app.reader import Div, Normalize, Transpose
from paddle_serving_app.reader import DBPostProcess, FilterBoxes, GetRotateCropImage, SortedBoxes
if sys.argv[1] == 'gpu':
from paddle_serving_server_gpu.web_service import WebService
elif sys.argv[1] == 'cpu':
from paddle_serving_server.web_service import WebService
import time
import re
import base64
class OCRService(WebService):
def init_rec(self):
self.ocr_reader = OCRReader()
def preprocess(self, feed=[], fetch=[]):
img_list = []
for feed_data in feed:
data = base64.b64decode(feed_data["image"].encode('utf8'))
data = np.fromstring(data, np.uint8)
im = cv2.imdecode(data, cv2.IMREAD_COLOR)
img_list.append(im)
max_wh_ratio = 0
for i, boximg in enumerate(img_list):
h, w = boximg.shape[0:2]
wh_ratio = w * 1.0 / h
max_wh_ratio = max(max_wh_ratio, wh_ratio)
_, w, h = self.ocr_reader.resize_norm_img(img_list[0],
max_wh_ratio).shape
imgs = np.zeros((len(img_list), 3, w, h)).astype('float32')
for i, img in enumerate(img_list):
norm_img = self.ocr_reader.resize_norm_img(img, max_wh_ratio)
imgs[i] = norm_img
feed = {"image": imgs.copy()}
fetch = ["ctc_greedy_decoder_0.tmp_0", "softmax_0.tmp_0"]
return feed, fetch
def postprocess(self, feed={}, fetch=[], fetch_map=None):
rec_res = self.ocr_reader.postprocess(fetch_map, with_score=True)
res_lst = []
for res in rec_res:
res_lst.append(res[0])
res = {"res": res_lst}
return res
ocr_service = OCRService(name="ocr")
ocr_service.load_model_config("ocr_rec_model")
ocr_service.init_rec()
if sys.argv[1] == 'gpu':
ocr_service.set_gpus("0")
ocr_service.prepare_server(workdir="workdir", port=9292, device="gpu", gpuid=0)
ocr_service.run_debugger_service(gpu=True)
elif sys.argv[1] == 'cpu':
ocr_service.prepare_server(workdir="workdir", port=9292, device="cpu")
ocr_service.run_debugger_service()
ocr_service.run_web_service()
deploy/pdserving/rec_web_server.py deleted 100644 → 0
View file @ bd314018
# 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 paddle_serving_client import Client
from paddle_serving_app.reader import OCRReader
import cv2
import sys
import numpy as np
import os
from paddle_serving_client import Client
from paddle_serving_app.reader import Sequential, URL2Image, ResizeByFactor
from paddle_serving_app.reader import Div, Normalize, Transpose
from paddle_serving_app.reader import DBPostProcess, FilterBoxes, GetRotateCropImage, SortedBoxes
if sys.argv[1] == 'gpu':
from paddle_serving_server_gpu.web_service import WebService
elif sys.argv[1] == 'cpu':
from paddle_serving_server.web_service import WebService
import time
import re
import base64
class OCRService(WebService):
def init_rec(self):
self.ocr_reader = OCRReader()
def preprocess(self, feed=[], fetch=[]):
# TODO: to handle batch rec images
img_list = []
for feed_data in feed:
data = base64.b64decode(feed_data["image"].encode('utf8'))
data = np.fromstring(data, np.uint8)
im = cv2.imdecode(data, cv2.IMREAD_COLOR)
img_list.append(im)
feed_list = []
max_wh_ratio = 0
for i, boximg in enumerate(img_list):
h, w = boximg.shape[0:2]
wh_ratio = w * 1.0 / h
max_wh_ratio = max(max_wh_ratio, wh_ratio)
for img in img_list:
norm_img = self.ocr_reader.resize_norm_img(img, max_wh_ratio)
feed = {"image": norm_img}
feed_list.append(feed)
fetch = ["ctc_greedy_decoder_0.tmp_0", "softmax_0.tmp_0"]
return feed_list, fetch
def postprocess(self, feed={}, fetch=[], fetch_map=None):
rec_res = self.ocr_reader.postprocess(fetch_map, with_score=True)
res_lst = []
for res in rec_res:
res_lst.append(res[0])
res = {"res": res_lst}
return res
ocr_service = OCRService(name="ocr")
ocr_service.load_model_config("ocr_rec_model")
ocr_service.init_rec()
if sys.argv[1] == 'gpu':
ocr_service.set_gpus("0")
ocr_service.prepare_server(workdir="workdir", port=9292, device="gpu", gpuid=0)
elif sys.argv[1] == 'cpu':
ocr_service.prepare_server(workdir="workdir", port=9292, device="cpu")
ocr_service.run_rpc_service()
ocr_service.run_web_service()
doc/doc_ch/algorithm_overview.md 100644 → 100755
View file @ 78d51971
<a name="算法介绍"></a> <a name="算法介绍"></a>
## 算法介绍 ## 算法介绍
本文给出了PaddleOCR已支持的文本检测算法和文本识别算法列表，以及每个算法在**英文公开数据集**上的模型和指标，主要用于算法简介和算法性能对比，更多包括中文在内的其他数据集上的模型请参考[PP-OCR v1.1 系列模型下载](./models_list.md)。 本文给出了PaddleOCR已支持的文本检测算法和文本识别算法列表，以及每个算法在**英文公开数据集**上的模型和指标，主要用于算法简介和算法性能对比，更多包括中文在内的其他数据集上的模型请参考[PP-OCR v2.0 系列模型下载](./models_list.md)。
- [1.文本检测算法](#文本检测算法) - [1.文本检测算法](#文本检测算法)
- [2.文本识别算法](#文本识别算法) - [2.文本识别算法](#文本识别算法)
...@@ -9,25 +9,25 @@ ...@@ -9,25 +9,25 @@
### 1.文本检测算法 ### 1.文本检测算法
PaddleOCR开源的文本检测算法列表： PaddleOCR开源的文本检测算法列表：
- [x] DB([paper](https://arxiv.org/abs/1911.08947))（ppocr推荐） - [x] DB([paper]( https://arxiv.org/abs/1911.08947) )（ppocr推荐）
- [x] EAST([paper](https://arxiv.org/abs/1704.03155)) - [x] EAST([paper](https://arxiv.org/abs/1704.03155))
- [x] SAST([paper](https://arxiv.org/abs/1908.05498)) - [x] SAST([paper](https://arxiv.org/abs/1908.05498))
在ICDAR2015文本检测公开数据集上，算法效果如下： 在ICDAR2015文本检测公开数据集上，算法效果如下：
|模型|骨干网络|precision|recall|Hmean|下载链接| |模型|骨干网络|precision|recall|Hmean|下载链接|
|-|-|-|-|-|-| | --- | --- | --- | --- | --- | --- |
|EAST|ResNet50_vd|88.18%|85.51%|86.82%|[下载链接](link)| |EAST|ResNet50_vd|88.76%|81.36%|84.90%|[下载链接](https://paddleocr.bj.bcebos.com/dygraph_v2.0/en/det_r50_vd_east_v2.0_train.tar)|
|EAST|MobileNetV3|81.67%|79.83%|80.74%|[下载链接](link)| |EAST|MobileNetV3|78.24%|79.15%|78.69%|[下载链接](https://paddleocr.bj.bcebos.com/dygraph_v2.0/en/det_mv3_east_v2.0_train.tar)|
|DB|ResNet50_vd|83.79%|80.65%|82.19%|[下载链接](link)| |DB|ResNet50_vd|86.41%|78.72%|82.38%|[下载链接](https://paddleocr.bj.bcebos.com/dygraph_v2.0/en/det_r50_vd_db_v2.0_train.tar)|
|DB|MobileNetV3|75.92%|73.18%|74.53%|[下载链接](link)| |DB|MobileNetV3|77.29%|73.08%|75.12%|[下载链接](https://paddleocr.bj.bcebos.com/dygraph_v2.0/en/det_mv3_db_v2.0_train.tar)|
|SAST|ResNet50_vd|92.18%|82.96%|87.33%|[下载链接](link))| |SAST|ResNet50_vd|91.83%|81.80%|86.52%|[下载链接](https://paddleocr.bj.bcebos.com/dygraph_v2.0/en/det_r50_vd_sast_icdar15_v2.0_train.tar)|
在Total-text文本检测公开数据集上，算法效果如下： 在Total-text文本检测公开数据集上，算法效果如下：
|模型|骨干网络|precision|recall|Hmean|下载链接| |模型|骨干网络|precision|recall|Hmean|下载链接|
|-|-|-|-|-|-| | --- | --- | --- | --- | --- | --- |
|SAST|ResNet50_vd|88.74%|79.80%|84.03%|[下载链接](link)| |SAST|ResNet50_vd|89.05%|76.80%|82.47%|[下载链接](https://paddleocr.bj.bcebos.com/dygraph_v2.0/en/det_r50_vd_sast_totaltext_v2.0_train.tar)|
**说明：** SAST模型训练额外加入了icdar2013、icdar2017、COCO-Text、ArT等公开数据集进行调优。PaddleOCR用到的经过整理格式的英文公开数据集下载：[百度云地址](https://pan.baidu.com/s/12cPnZcVuV1zn5DOd4mqjVw) (提取码: 2bpi) **说明：** SAST模型训练额外加入了icdar2013、icdar2017、COCO-Text、ArT等公开数据集进行调优。PaddleOCR用到的经过整理格式的英文公开数据集下载：[百度云地址](https://pan.baidu.com/s/12cPnZcVuV1zn5DOd4mqjVw) (提取码: 2bpi)
...@@ -38,9 +38,9 @@ PaddleOCR文本检测算法的训练和使用请参考文档教程中[模型训 ...@@ -38,9 +38,9 @@ PaddleOCR文本检测算法的训练和使用请参考文档教程中[模型训
### 2.文本识别算法 ### 2.文本识别算法
PaddleOCR基于动态图开源的文本识别算法列表： PaddleOCR基于动态图开源的文本识别算法列表：
- [x] CRNN([paper](https://arxiv.org/abs/1507.05717))（ppocr推荐） - [x] CRNN([paper](https://arxiv.org/abs/1507.05717) )（ppocr推荐）
- [x] Rosetta([paper](https://arxiv.org/abs/1910.05085)) - [x] Rosetta([paper](https://arxiv.org/abs/1910.05085))
- [x] STAR-Net([paper](http://www.bmva.org/bmvc/2016/papers/paper043/index.html)) - [ ] STAR-Net([paper](http://www.bmva.org/bmvc/2016/papers/paper043/index.html)) coming soon
- [ ] RARE([paper](https://arxiv.org/abs/1603.03915v1)) coming soon - [ ] RARE([paper](https://arxiv.org/abs/1603.03915v1)) coming soon
- [ ] SRN([paper](https://arxiv.org/abs/2003.12294)) coming soon - [ ] SRN([paper](https://arxiv.org/abs/2003.12294)) coming soon
...@@ -48,12 +48,9 @@ PaddleOCR基于动态图开源的文本识别算法列表： ...@@ -48,12 +48,9 @@ PaddleOCR基于动态图开源的文本识别算法列表：
|模型|骨干网络|Avg Accuracy|模型存储命名|下载链接| |模型|骨干网络|Avg Accuracy|模型存储命名|下载链接|
|-|-|-|-|-| |-|-|-|-|-|
|Rosetta|Resnet34_vd|80.24%|rec_r34_vd_none_none_ctc|[下载链接](link)| |Rosetta|Resnet34_vd|80.9%|rec_r34_vd_none_none_ctc|[下载链接](https://paddleocr.bj.bcebos.com/dygraph_v2.0/en/rec_r34_vd_none_none_ctc_v2.0_train.tar)|
|Rosetta|MobileNetV3|78.16%|rec_mv3_none_none_ctc|[下载链接](link)| |Rosetta|MobileNetV3|78.05%|rec_mv3_none_none_ctc|[下载链接](https://paddleocr.bj.bcebos.com/dygraph_v2.0/en/rec_mv3_none_none_ctc_v2.0_train.tar)|
|CRNN|Resnet34_vd|82.20%|rec_r34_vd_none_bilstm_ctc|[下载链接](link)| |CRNN|Resnet34_vd|82.76%|rec_r34_vd_none_bilstm_ctc|[下载链接](https://paddleocr.bj.bcebos.com/dygraph_v2.0/en/rec_r34_vd_none_bilstm_ctc_v2.0_train.tar)|
|CRNN|MobileNetV3|79.37%|rec_mv3_none_bilstm_ctc|[下载链接](link)| |CRNN|MobileNetV3|79.97%|rec_mv3_none_bilstm_ctc|[下载链接](https://paddleocr.bj.bcebos.com/dygraph_v2.0/en/rec_mv3_none_bilstm_ctc_v2.0_train.tar)|
|STAR-Net|Resnet34_vd|83.93%|rec_r34_vd_tps_bilstm_ctc|[下载链接](link)|
|STAR-Net|MobileNetV3|81.56%|rec_mv3_tps_bilstm_ctc|[下载链接](link)|
PaddleOCR文本识别算法的训练和使用请参考文档教程中[模型训练/评估中的文本识别部分](./recognition.md)。 PaddleOCR文本识别算法的训练和使用请参考文档教程中[模型训练/评估中的文本识别部分](./recognition.md)。
doc/doc_ch/angle_class.md
View file @ 78d51971
...@@ -62,9 +62,9 @@ PaddleOCR提供了训练脚本、评估脚本和预测脚本。 ...@@ -62,9 +62,9 @@ PaddleOCR提供了训练脚本、评估脚本和预测脚本。
*如果您安装的是cpu版本，请将配置文件中的 `use_gpu` 字段修改为false* *如果您安装的是cpu版本，请将配置文件中的 `use_gpu` 字段修改为false*
``` ```
# GPU训练 支持单卡，多卡训练，通过selected_gpus指定卡号 # GPU训练 支持单卡，多卡训练，通过 '--gpus' 指定卡号，如果使用的paddle版本小于2.0rc1，请使用'--select_gpus'参数选择要使用的GPU
# 启动训练，下面的命令已经写入train.sh文件中，只需修改文件里的配置文件路径即可 # 启动训练，下面的命令已经写入train.sh文件中，只需修改文件里的配置文件路径即可
python3 -m paddle.distributed.launch --selected_gpus '0,1,2,3,4,5,6,7' tools/train.py -c configs/cls/cls_mv3.yml python3 -m paddle.distributed.launch --gpus '0,1,2,3,4,5,6,7' tools/train.py -c configs/cls/cls_mv3.yml
``` ```
- 数据增强 - 数据增强
...@@ -74,7 +74,7 @@ PaddleOCR提供了多种数据增强方式，如果您希望在训练时加入 ...@@ -74,7 +74,7 @@ PaddleOCR提供了多种数据增强方式，如果您希望在训练时加入
默认的扰动方式有：颜色空间转换(cvtColor)、模糊(blur)、抖动(jitter)、噪声(Gasuss noise)、随机切割(random crop)、透视(perspective)、颜色反转(reverse),随机数据增强(RandAugment)。 默认的扰动方式有：颜色空间转换(cvtColor)、模糊(blur)、抖动(jitter)、噪声(Gasuss noise)、随机切割(random crop)、透视(perspective)、颜色反转(reverse),随机数据增强(RandAugment)。
训练过程中除随机数据增强外每种扰动方式以50%的概率被选择，具体代码实现请参考： 训练过程中除随机数据增强外每种扰动方式以50%的概率被选择，具体代码实现请参考：
[rec_img_aug.py](../../ppocr/data/imaug/rec_img_aug.py) [rec_img_aug.py](../../ppocr/data/imaug/rec_img_aug.py)
[randaugment.py](../../ppocr/data/imaug/randaugment.py) [randaugment.py](../../ppocr/data/imaug/randaugment.py)
*由于OpenCV的兼容性问题，扰动操作暂时只支持linux* *由于OpenCV的兼容性问题，扰动操作暂时只支持linux*
......
doc/doc_ch/detection.md
View file @ 78d51971
...@@ -76,8 +76,8 @@ tar -xf ./pretrain_models/MobileNetV3_large_x0_5_pretrained.tar ./pretrain_model ...@@ -76,8 +76,8 @@ tar -xf ./pretrain_models/MobileNetV3_large_x0_5_pretrained.tar ./pretrain_model
# 单机单卡训练 mv3_db 模型 # 单机单卡训练 mv3_db 模型
python3 tools/train.py -c configs/det/det_mv3_db.yml \ python3 tools/train.py -c configs/det/det_mv3_db.yml \
-o Global.pretrain_weights=./pretrain_models/MobileNetV3_large_x0_5_pretrained/ -o Global.pretrain_weights=./pretrain_models/MobileNetV3_large_x0_5_pretrained/
# 单机多卡训练，通过 --select_gpus 参数设置使用的GPU ID； # 单机多卡训练，通过 --gpus 参数设置使用的GPU ID；如果使用的paddle版本小于2.0rc1，请使用'--select_gpus'参数选择要使用的GPU
python3 -m paddle.distributed.launch --selected_gpus '0,1,2,3' tools/train.py -c configs/det/det_mv3_db.yml \ python3 -m paddle.distributed.launch --gpus '0,1,2,3' tools/train.py -c configs/det/det_mv3_db.yml \
-o Global.pretrain_weights=./pretrain_models/MobileNetV3_large_x0_5_pretrained/ -o Global.pretrain_weights=./pretrain_models/MobileNetV3_large_x0_5_pretrained/
``` ```
...@@ -107,17 +107,13 @@ PaddleOCR计算三个OCR检测相关的指标，分别是：Precision、Recall ...@@ -107,17 +107,13 @@ PaddleOCR计算三个OCR检测相关的指标，分别是：Precision、Recall
运行如下代码，根据配置文件`det_db_mv3.yml`中`save_res_path`指定的测试集检测结果文件，计算评估指标。 运行如下代码，根据配置文件`det_db_mv3.yml`中`save_res_path`指定的测试集检测结果文件，计算评估指标。
评估时设置后处理参数`box_thresh=0.6`，`unclip_ratio=1.5`，使用不同数据集、不同模型训练，可调整这两个参数进行优化 评估时设置后处理参数`box_thresh=0.5`，`unclip_ratio=1.5`，使用不同数据集、不同模型训练，可调整这两个参数进行优化
```shell
python3 tools/eval.py -c configs/det/det_mv3_db.yml -o Global.checkpoints="{path/to/weights}/best_accuracy" PostProcess.box_thresh=0.6 PostProcess.unclip_ratio=1.5
```
训练中模型参数默认保存在`Global.save_model_dir`目录下。在评估指标时，需要设置`Global.checkpoints`指向保存的参数文件。 训练中模型参数默认保存在`Global.save_model_dir`目录下。在评估指标时，需要设置`Global.checkpoints`指向保存的参数文件。
比如：
```shell ```shell
python3 tools/eval.py -c configs/det/det_mv3_db.yml -o Global.checkpoints="./output/det_db/best_accuracy" PostProcess.box_thresh=0.6 PostProcess.unclip_ratio=1.5 python3 tools/eval.py -c configs/det/det_mv3_db.yml -o Global.checkpoints="{path/to/weights}/best_accuracy" PostProcess.box_thresh=0.5 PostProcess.unclip_ratio=1.5
``` ```
* 注：`box_thresh`、`unclip_ratio`是DB后处理所需要的参数，在评估EAST模型时不需要设置 * 注：`box_thresh`、`unclip_ratio`是DB后处理所需要的参数，在评估EAST模型时不需要设置
## 测试检测效果 ## 测试检测效果
......
doc/doc_ch/inference.md 100644 → 100755
View file @ 78d51971
...@@ -22,9 +22,8 @@ inference 模型（`paddle.jit.save`保存的模型） ...@@ -22,9 +22,8 @@ inference 模型（`paddle.jit.save`保存的模型）
- [三、文本识别模型推理](#文本识别模型推理) - [三、文本识别模型推理](#文本识别模型推理)
- [1. 超轻量中文识别模型推理](#超轻量中文识别模型推理) - [1. 超轻量中文识别模型推理](#超轻量中文识别模型推理)
- [2. 基于CTC损失的识别模型推理](#基于CTC损失的识别模型推理) - [2. 基于CTC损失的识别模型推理](#基于CTC损失的识别模型推理)
- [3. 基于Attention损失的识别模型推理](#基于Attention损失的识别模型推理) - [3. 自定义文本识别字典的推理](#自定义文本识别字典的推理)
- [4. 自定义文本识别字典的推理](#自定义文本识别字典的推理) - [4. 多语言模型的推理](#多语言模型的推理)
- [5. 多语言模型的推理](#多语言模型的推理)
- [四、方向分类模型推理](#方向识别模型推理) - [四、方向分类模型推理](#方向识别模型推理)
- [1. 方向分类模型推理](#方向分类模型推理) - [1. 方向分类模型推理](#方向分类模型推理)
...@@ -129,24 +128,32 @@ python3 tools/export_model.py -c configs/cls/cls_mv3.yml -o Global.pretrained_mo ...@@ -129,24 +128,32 @@ python3 tools/export_model.py -c configs/cls/cls_mv3.yml -o Global.pretrained_mo
超轻量中文检测模型推理，可以执行如下命令： 超轻量中文检测模型推理，可以执行如下命令：
``` ```
python3 tools/infer/predict_det.py --image_dir="./doc/imgs/2.jpg" --det_model_dir="./inference/det_db/" # 下载超轻量中文检测模型：
wget https://paddleocr.bj.bcebos.com/dygraph_v2.0/ch/ch_ppocr_mobile_v2.0_det_infer.tar
tar xf ch_ppocr_mobile_v2.0_det_infer.tar
python3 tools/infer/predict_det.py --image_dir="./doc/imgs/22.jpg" --det_model_dir="./ch_ppocr_mobile_v2.0_det_infer/"
``` ```
可视化文本检测结果默认保存到`./inference_results`文件夹里面，结果文件的名称前缀为'det_res'。结果示例如下： 可视化文本检测结果默认保存到`./inference_results`文件夹里面，结果文件的名称前缀为'det_res'。结果示例如下：
![](../imgs_results/det_res_2.jpg) ![](../imgs_results/det_res_22.jpg)
通过参数`limit_type`和`det_limit_side_len`来对图片的尺寸进行限制限，`limit_type=max`为限制长边长度<`det_limit_side_len`，`limit_type=min`为限制短边长度>`det_limit_side_len`, 通过参数`limit_type`和`det_limit_side_len`来对图片的尺寸进行限制，
图片不满足限制条件时(`limit_type=max`时长边长度>`det_limit_side_len`或`limit_type=min`时短边长度<`det_limit_side_len`)，将对图片进行等比例缩放。 `litmit_type`可选参数为[`max`, `min`]，
该参数默认设置为`limit_type='max',det_max_side_len=960`。 如果输入图片的分辨率比较大，而且想使用更大的分辨率预测，可以执行如下命令： `det_limit_size_len` 为正整数，一般设置为32 的倍数，比如960。
参数默认设置为`limit_type='max', det_limit_side_len=960`。表示网络输入图像的最长边不能超过960，
如果超过这个值，会对图像做等宽比的resize操作，确保最长边为`det_limit_side_len`。
设置为`limit_type='min', det_limit_side_len=960` 则表示限制图像的最短边为960。
如果输入图片的分辨率比较大，而且想使用更大的分辨率预测，可以设置det_limit_side_len 为想要的值，比如1216：
``` ```
python3 tools/infer/predict_det.py --image_dir="./doc/imgs/2.jpg" --det_model_dir="./inference/det_db/" --det_limit_type=max --det_limit_side_len=1200 python3 tools/infer/predict_det.py --image_dir="./doc/imgs/2.jpg" --det_model_dir="./inference/det_db/" --det_limit_type=max --det_limit_side_len=1216
``` ```
如果想使用CPU进行预测，执行命令如下 如果想使用CPU进行预测，执行命令如下
``` ```
python3 tools/infer/predict_det.py --image_dir="./doc/imgs/2.jpg" --det_model_dir="./inference/det_db/" --use_gpu=False python3 tools/infer/predict_det.py --image_dir="./doc/imgs/2.jpg" --det_model_dir="./inference/det_db/" --use_gpu=False
``` ```
<a name="DB文本检测模型推理"></a> <a name="DB文本检测模型推理"></a>
...@@ -173,7 +180,7 @@ python3 tools/infer/predict_det.py --image_dir="./doc/imgs_en/img_10.jpg" --det_ ...@@ -173,7 +180,7 @@ python3 tools/infer/predict_det.py --image_dir="./doc/imgs_en/img_10.jpg" --det_
<a name="EAST文本检测模型推理"></a> <a name="EAST文本检测模型推理"></a>
### 3. EAST文本检测模型推理 ### 3. EAST文本检测模型推理
首先将EAST文本检测训练过程中保存的模型，转换成inference model。以基于Resnet50_vd骨干网络，在ICDAR2015英文数据集训练的模型为例（ [模型下载地址 (coming soon)](link) )，可以使用如下命令进行转换： 首先将EAST文本检测训练过程中保存的模型，转换成inference model。以基于Resnet50_vd骨干网络，在ICDAR2015英文数据集训练的模型为例（ [模型下载地址](https://paddleocr.bj.bcebos.com/dygraph_v2.0/en/det_r50_vd_east_v2.0_train.tar) )，可以使用如下命令进行转换：
``` ```
python3 tools/export_model.py -c configs/det/det_r50_vd_east.yml -o Global.pretrained_model=./det_r50_vd_east_v2.0_train/best_accuracy Global.load_static_weights=False Global.save_inference_dir=./inference/det_east python3 tools/export_model.py -c configs/det/det_r50_vd_east.yml -o Global.pretrained_model=./det_r50_vd_east_v2.0_train/best_accuracy Global.load_static_weights=False Global.save_inference_dir=./inference/det_east
...@@ -186,7 +193,7 @@ python3 tools/infer/predict_det.py --det_algorithm="EAST" --image_dir="./doc/img ...@@ -186,7 +193,7 @@ python3 tools/infer/predict_det.py --det_algorithm="EAST" --image_dir="./doc/img
``` ```
可视化文本检测结果默认保存到`./inference_results`文件夹里面，结果文件的名称前缀为'det_res'。结果示例如下： 可视化文本检测结果默认保存到`./inference_results`文件夹里面，结果文件的名称前缀为'det_res'。结果示例如下：
(coming soon) ![](../imgs_results/det_res_img_10_east.jpg)
**注意**：本代码库中，EAST后处理Locality-Aware NMS有python和c++两种版本，c++版速度明显快于python版。由于c++版本nms编译版本问题，只有python3.5环境下会调用c++版nms，其他情况将调用python版nms。 **注意**：本代码库中，EAST后处理Locality-Aware NMS有python和c++两种版本，c++版速度明显快于python版。由于c++版本nms编译版本问题，只有python3.5环境下会调用c++版nms，其他情况将调用python版nms。
...@@ -194,7 +201,7 @@ python3 tools/infer/predict_det.py --det_algorithm="EAST" --image_dir="./doc/img ...@@ -194,7 +201,7 @@ python3 tools/infer/predict_det.py --det_algorithm="EAST" --image_dir="./doc/img
<a name="SAST文本检测模型推理"></a> <a name="SAST文本检测模型推理"></a>
### 4. SAST文本检测模型推理 ### 4. SAST文本检测模型推理
#### (1). 四边形文本检测模型（ICDAR2015） #### (1). 四边形文本检测模型（ICDAR2015）
首先将SAST文本检测训练过程中保存的模型，转换成inference model。以基于Resnet50_vd骨干网络，在ICDAR2015英文数据集训练的模型为例([模型下载地址(coming soon)](link))，可以使用如下命令进行转换： 首先将SAST文本检测训练过程中保存的模型，转换成inference model。以基于Resnet50_vd骨干网络，在ICDAR2015英文数据集训练的模型为例([模型下载地址](https://paddleocr.bj.bcebos.com/dygraph_v2.0/en/det_r50_vd_sast_icdar15_v2.0_train.tar))，可以使用如下命令进行转换：
``` ```
python3 tools/export_model.py -c configs/det/det_r50_vd_sast_icdar15.yml -o Global.pretrained_model=./det_r50_vd_sast_icdar15_v2.0_train/best_accuracy Global.load_static_weights=False Global.save_inference_dir=./inference/det_sast_ic15 python3 tools/export_model.py -c configs/det/det_r50_vd_sast_icdar15.yml -o Global.pretrained_model=./det_r50_vd_sast_icdar15_v2.0_train/best_accuracy Global.load_static_weights=False Global.save_inference_dir=./inference/det_sast_ic15
...@@ -205,10 +212,10 @@ python3 tools/infer/predict_det.py --det_algorithm="SAST" --image_dir="./doc/img ...@@ -205,10 +212,10 @@ python3 tools/infer/predict_det.py --det_algorithm="SAST" --image_dir="./doc/img
``` ```
可视化文本检测结果默认保存到`./inference_results`文件夹里面，结果文件的名称前缀为'det_res'。结果示例如下： 可视化文本检测结果默认保存到`./inference_results`文件夹里面，结果文件的名称前缀为'det_res'。结果示例如下：
(coming soon) ![](../imgs_results/det_res_img_10_sast.jpg)
#### (2). 弯曲文本检测模型（Total-Text） #### (2). 弯曲文本检测模型（Total-Text）
首先将SAST文本检测训练过程中保存的模型，转换成inference model。以基于Resnet50_vd骨干网络，在Total-Text英文数据集训练的模型为例（[模型下载地址(coming soon)](link))，可以使用如下命令进行转换： 首先将SAST文本检测训练过程中保存的模型，转换成inference model。以基于Resnet50_vd骨干网络，在Total-Text英文数据集训练的模型为例（[模型下载地址](https://paddleocr.bj.bcebos.com/dygraph_v2.0/en/det_r50_vd_sast_totaltext_v2.0_train.tar))，可以使用如下命令进行转换：
``` ```
python3 tools/export_model.py -c configs/det/det_r50_vd_sast_totaltext.yml -o Global.pretrained_model=./det_r50_vd_sast_totaltext_v2.0_train/best_accuracy Global.load_static_weights=False Global.save_inference_dir=./inference/det_sast_tt python3 tools/export_model.py -c configs/det/det_r50_vd_sast_totaltext.yml -o Global.pretrained_model=./det_r50_vd_sast_totaltext_v2.0_train/best_accuracy Global.load_static_weights=False Global.save_inference_dir=./inference/det_sast_tt
...@@ -221,7 +228,7 @@ python3 tools/infer/predict_det.py --det_algorithm="SAST" --image_dir="./doc/img ...@@ -221,7 +228,7 @@ python3 tools/infer/predict_det.py --det_algorithm="SAST" --image_dir="./doc/img
``` ```
可视化文本检测结果默认保存到`./inference_results`文件夹里面，结果文件的名称前缀为'det_res'。结果示例如下： 可视化文本检测结果默认保存到`./inference_results`文件夹里面，结果文件的名称前缀为'det_res'。结果示例如下：
(coming soon) ![](../imgs_results/det_res_img623_sast.jpg)
**注意**：本代码库中，SAST后处理Locality-Aware NMS有python和c++两种版本，c++版速度明显快于python版。由于c++版本nms编译版本问题，只有python3.5环境下会调用c++版nms，其他情况将调用python版nms。 **注意**：本代码库中，SAST后处理Locality-Aware NMS有python和c++两种版本，c++版速度明显快于python版。由于c++版本nms编译版本问题，只有python3.5环境下会调用c++版nms，其他情况将调用python版nms。
...@@ -268,16 +275,6 @@ CRNN 文本识别模型推理，可以执行如下命令： ...@@ -268,16 +275,6 @@ CRNN 文本识别模型推理，可以执行如下命令：
python3 tools/infer/predict_rec.py --image_dir="./doc/imgs_words_en/word_336.png" --rec_model_dir="./inference/rec_crnn/" --rec_image_shape="3, 32, 100" --rec_char_type="en" python3 tools/infer/predict_rec.py --image_dir="./doc/imgs_words_en/word_336.png" --rec_model_dir="./inference/rec_crnn/" --rec_image_shape="3, 32, 100" --rec_char_type="en"
``` ```
<a name="基于Attention损失的识别模型推理"></a>
### 3. 基于Attention损失的识别模型推理
基于Attention损失的识别模型与ctc不同，需要额外设置识别算法参数 --rec_algorithm="RARE"
RARE 文本识别模型推理，可以执行如下命令：
```
python3 tools/infer/predict_rec.py --image_dir="./doc/imgs_words_en/word_336.png" --rec_model_dir="./inference/rare/" --rec_image_shape="3, 32, 100" --rec_char_type="en" --rec_algorithm="RARE"
```
![](../imgs_words_en/word_336.png) ![](../imgs_words_en/word_336.png)
执行命令后，上面图像的识别结果如下： 执行命令后，上面图像的识别结果如下：
...@@ -297,7 +294,7 @@ self.character_str = "0123456789abcdefghijklmnopqrstuvwxyz" ...@@ -297,7 +294,7 @@ self.character_str = "0123456789abcdefghijklmnopqrstuvwxyz"
dict_character = list(self.character_str) dict_character = list(self.character_str)
``` ```
### 4. 自定义文本识别字典的推理 ### 3. 自定义文本识别字典的推理
如果训练时修改了文本的字典，在使用inference模型预测时，需要通过`--rec_char_dict_path`指定使用的字典路径，并且设置 `rec_char_type=ch` 如果训练时修改了文本的字典，在使用inference模型预测时，需要通过`--rec_char_dict_path`指定使用的字典路径，并且设置 `rec_char_type=ch`
``` ```
...@@ -305,7 +302,7 @@ python3 tools/infer/predict_rec.py --image_dir="./doc/imgs_words_en/word_336.png ...@@ -305,7 +302,7 @@ python3 tools/infer/predict_rec.py --image_dir="./doc/imgs_words_en/word_336.png
``` ```
<a name="多语言模型的推理"></a> <a name="多语言模型的推理"></a>
### 5. 多语言模型的推理 ### 4. 多语言模型的推理
如果您需要预测的是其他语言模型，在使用inference模型预测时，需要通过`--rec_char_dict_path`指定使用的字典路径, 同时为了得到正确的可视化结果， 如果您需要预测的是其他语言模型，在使用inference模型预测时，需要通过`--rec_char_dict_path`指定使用的字典路径, 同时为了得到正确的可视化结果，
需要通过 `--vis_font_path` 指定可视化的字体路径，`doc/` 路径下有默认提供的小语种字体，例如韩文识别： 需要通过 `--vis_font_path` 指定可视化的字体路径，`doc/` 路径下有默认提供的小语种字体，例如韩文识别：
......
doc/doc_ch/installation.md
View file @ 78d51971
...@@ -2,7 +2,7 @@ ...@@ -2,7 +2,7 @@
经测试PaddleOCR可在glibc 2.23上运行，您也可以测试其他glibc版本或安装glic 2.23 经测试PaddleOCR可在glibc 2.23上运行，您也可以测试其他glibc版本或安装glic 2.23
PaddleOCR 工作环境 PaddleOCR 工作环境
- PaddlePaddle 2.0rc0+ ，推荐使用 PaddlePaddle 2.0rc0 - PaddlePaddle 1.8+ ，推荐使用 PaddlePaddle 2.0rc1
- python3.7 - python3.7
- glibc 2.23 - glibc 2.23
- cuDNN 7.6+ (GPU) - cuDNN 7.6+ (GPU)
...@@ -35,11 +35,11 @@ sudo docker container exec -it ppocr /bin/bash ...@@ -35,11 +35,11 @@ sudo docker container exec -it ppocr /bin/bash
pip3 install --upgrade pip pip3 install --upgrade pip
如果您的机器安装的是CUDA9或CUDA10，请运行以下命令安装 如果您的机器安装的是CUDA9或CUDA10，请运行以下命令安装
python3 -m pip install paddlepaddle-gpu==2.0.0rc0 -i https://mirror.baidu.com/pypi/simple python3 -m pip install paddlepaddle-gpu==2.0.0rc1 -i https://mirror.baidu.com/pypi/simple
如果您的机器是CPU，请运行以下命令安装 如果您的机器是CPU，请运行以下命令安装
python3 -m pip install paddlepaddle==2.0.0rc0 -i https://mirror.baidu.com/pypi/simple python3 -m pip install paddlepaddle==2.0.0rc1 -i https://mirror.baidu.com/pypi/simple
更多的版本需求，请参照[安装文档](https://www.paddlepaddle.org.cn/install/quick)中的说明进行操作。 更多的版本需求，请参照[安装文档](https://www.paddlepaddle.org.cn/install/quick)中的说明进行操作。
``` ```
......
doc/doc_ch/models_list.md
View file @ 78d51971
## OCR模型列表（V2.0，2020年12月12日更新） ## OCR模型列表（V2.0，2020年12月12日更新）
**说明** ：2.0版模型和[1.1版模型](https://github.com/PaddlePaddle/PaddleOCR/blob/develop/doc/doc_ch/models_list.md)的主要区别在于动态图训练vs.静态图训练，模型性能上无明显差距。
- [一、文本检测模型](#文本检测模型) - [一、文本检测模型](#文本检测模型)
- [二、文本识别模型](#文本识别模型) - [二、文本识别模型](#文本识别模型)
...@@ -21,7 +22,7 @@ PaddleOCR提供的可下载模型包括`推理模型`、`训练模型`、`预训 ...@@ -21,7 +22,7 @@ PaddleOCR提供的可下载模型包括`推理模型`、`训练模型`、`预训
|模型名称|模型简介|配置文件|推理模型大小|下载地址| |模型名称|模型简介|配置文件|推理模型大小|下载地址|
| --- | --- | --- | --- | --- | | --- | --- | --- | --- | --- |
|ch_ppocr_mobile_slim_v2.0_det|slim裁剪版超轻量模型，支持中英文、多语种文本检测|[ch_det_mv3_db_v2.0.yml](../../configs/det/ch_ppocr_v2.0/ch_det_mv3_db_v2.0.yml)| |[推理模型 (coming soon)](link) / [slim模型 (coming soon)](link)| |ch_ppocr_mobile_slim_v2.0_det|slim裁剪版超轻量模型，支持中英文、多语种文本检测|[ch_det_mv3_db_v2.0.yml](../../configs/det/ch_ppocr_v2.0/ch_det_mv3_db_v2.0.yml)| |推理模型 (coming soon) / slim模型 (coming soon)|
|ch_ppocr_mobile_v2.0_det|原始超轻量模型，支持中英文、多语种文本检测|[ch_det_mv3_db_v2.0.yml](../../configs/det/ch_ppocr_v2.0/ch_det_mv3_db_v2.0.yml)|3M|[推理模型](https://paddleocr.bj.bcebos.com/dygraph_v2.0/ch/ch_ppocr_mobile_v2.0_det_infer.tar) / [训练模型](https://paddleocr.bj.bcebos.com/dygraph_v2.0/ch/ch_ppocr_mobile_v2.0_det_train.tar)| |ch_ppocr_mobile_v2.0_det|原始超轻量模型，支持中英文、多语种文本检测|[ch_det_mv3_db_v2.0.yml](../../configs/det/ch_ppocr_v2.0/ch_det_mv3_db_v2.0.yml)|3M|[推理模型](https://paddleocr.bj.bcebos.com/dygraph_v2.0/ch/ch_ppocr_mobile_v2.0_det_infer.tar) / [训练模型](https://paddleocr.bj.bcebos.com/dygraph_v2.0/ch/ch_ppocr_mobile_v2.0_det_train.tar)|
|ch_ppocr_server_v2.0_det|通用模型，支持中英文、多语种文本检测，比超轻量模型更大，但效果更好|[ch_det_res18_db_v2.0.yml](../../configs/det/ch_ppocr_v2.0/ch_det_res18_db_v2.0.yml)|47M|[推理模型](https://paddleocr.bj.bcebos.com/dygraph_v2.0/ch/ch_ppocr_server_v2.0_det_infer.tar) / [训练模型](https://paddleocr.bj.bcebos.com/dygraph_v2.0/ch/ch_ppocr_server_v2.0_det_train.tar)| |ch_ppocr_server_v2.0_det|通用模型，支持中英文、多语种文本检测，比超轻量模型更大，但效果更好|[ch_det_res18_db_v2.0.yml](../../configs/det/ch_ppocr_v2.0/ch_det_res18_db_v2.0.yml)|47M|[推理模型](https://paddleocr.bj.bcebos.com/dygraph_v2.0/ch/ch_ppocr_server_v2.0_det_infer.tar) / [训练模型](https://paddleocr.bj.bcebos.com/dygraph_v2.0/ch/ch_ppocr_server_v2.0_det_train.tar)|
...@@ -34,7 +35,7 @@ PaddleOCR提供的可下载模型包括`推理模型`、`训练模型`、`预训 ...@@ -34,7 +35,7 @@ PaddleOCR提供的可下载模型包括`推理模型`、`训练模型`、`预训
|模型名称|模型简介|配置文件|推理模型大小|下载地址| |模型名称|模型简介|配置文件|推理模型大小|下载地址|
| --- | --- | --- | --- | --- | | --- | --- | --- | --- | --- |
|ch_ppocr_mobile_slim_v2.0_rec|slim裁剪量化版超轻量模型，支持中英文、数字识别|[rec_chinese_lite_train_v2.0.yml](../../configs/rec/ch_ppocr_v2.0/rec_chinese_lite_train_v2.0.yml)| |[推理模型 (coming soon)](link) / [slim模型 (coming soon)](link) | |ch_ppocr_mobile_slim_v2.0_rec|slim裁剪量化版超轻量模型，支持中英文、数字识别|[rec_chinese_lite_train_v2.0.yml](../../configs/rec/ch_ppocr_v2.0/rec_chinese_lite_train_v2.0.yml)| |推理模型 (coming soon) / slim模型 (coming soon) |
|ch_ppocr_mobile_v2.0_rec|原始超轻量模型，支持中英文、数字识别|[rec_chinese_lite_train_v2.0.yml](../../configs/rec/ch_ppocr_v2.0/rec_chinese_lite_train_v2.0.yml)|3.71M|[推理模型](https://paddleocr.bj.bcebos.com/dygraph_v2.0/ch/ch_ppocr_mobile_v2.0_rec_infer.tar) / [训练模型](https://paddleocr.bj.bcebos.com/dygraph_v2.0/ch/ch_ppocr_mobile_v2.0_rec_train.tar) / [预训练模型](https://paddleocr.bj.bcebos.com/dygraph_v2.0/ch/ch_ppocr_mobile_v2.0_rec_pre.tar) | |ch_ppocr_mobile_v2.0_rec|原始超轻量模型，支持中英文、数字识别|[rec_chinese_lite_train_v2.0.yml](../../configs/rec/ch_ppocr_v2.0/rec_chinese_lite_train_v2.0.yml)|3.71M|[推理模型](https://paddleocr.bj.bcebos.com/dygraph_v2.0/ch/ch_ppocr_mobile_v2.0_rec_infer.tar) / [训练模型](https://paddleocr.bj.bcebos.com/dygraph_v2.0/ch/ch_ppocr_mobile_v2.0_rec_train.tar) / [预训练模型](https://paddleocr.bj.bcebos.com/dygraph_v2.0/ch/ch_ppocr_mobile_v2.0_rec_pre.tar) |
|ch_ppocr_server_v2.0_rec|通用模型，支持中英文、数字识别|[rec_chinese_common_train_v2.0.yml](../../configs/rec/ch_ppocr_v2.0/rec_chinese_common_train_v2.0.yml)|94.8M|[推理模型](https://paddleocr.bj.bcebos.com/dygraph_v2.0/ch/ch_ppocr_server_v2.0_rec_infer.tar) / [训练模型](https://paddleocr.bj.bcebos.com/dygraph_v2.0/ch/ch_ppocr_server_v2.0_rec_train.tar) / [预训练模型](https://paddleocr.bj.bcebos.com/dygraph_v2.0/ch/ch_ppocr_server_v2.0_rec_pre.tar) | |ch_ppocr_server_v2.0_rec|通用模型，支持中英文、数字识别|[rec_chinese_common_train_v2.0.yml](../../configs/rec/ch_ppocr_v2.0/rec_chinese_common_train_v2.0.yml)|94.8M|[推理模型](https://paddleocr.bj.bcebos.com/dygraph_v2.0/ch/ch_ppocr_server_v2.0_rec_infer.tar) / [训练模型](https://paddleocr.bj.bcebos.com/dygraph_v2.0/ch/ch_ppocr_server_v2.0_rec_train.tar) / [预训练模型](https://paddleocr.bj.bcebos.com/dygraph_v2.0/ch/ch_ppocr_server_v2.0_rec_pre.tar) |
...@@ -45,7 +46,7 @@ PaddleOCR提供的可下载模型包括`推理模型`、`训练模型`、`预训 ...@@ -45,7 +46,7 @@ PaddleOCR提供的可下载模型包括`推理模型`、`训练模型`、`预训
|模型名称|模型简介|配置文件|推理模型大小|下载地址| |模型名称|模型简介|配置文件|推理模型大小|下载地址|
| --- | --- | --- | --- | --- | | --- | --- | --- | --- | --- |
|en_number_mobile_slim_v2.0_rec|slim裁剪量化版超轻量模型，支持英文、数字识别|[rec_en_number_lite_train.yml](../../configs/rec/multi_language/rec_en_number_lite_train.yml)| |[推理模型 (coming soon )](link) / [slim模型 (coming soon)](link) | |en_number_mobile_slim_v2.0_rec|slim裁剪量化版超轻量模型，支持英文、数字识别|[rec_en_number_lite_train.yml](../../configs/rec/multi_language/rec_en_number_lite_train.yml)| | 推理模型 (coming soon) / slim模型 (coming soon) |
|en_number_mobile_v2.0_rec|原始超轻量模型，支持英文、数字识别|[rec_en_number_lite_train.yml](../../configs/rec/multi_language/rec_en_number_lite_train.yml)|2.56M|[推理模型](https://paddleocr.bj.bcebos.com/dygraph_v2.0/multilingual/en_number_mobile_v2.0_rec_infer.tar) / [训练模型](https://paddleocr.bj.bcebos.com/dygraph_v2.0/multilingual/en_number_mobile_v2.0_rec_train.tar) | |en_number_mobile_v2.0_rec|原始超轻量模型，支持英文、数字识别|[rec_en_number_lite_train.yml](../../configs/rec/multi_language/rec_en_number_lite_train.yml)|2.56M|[推理模型](https://paddleocr.bj.bcebos.com/dygraph_v2.0/multilingual/en_number_mobile_v2.0_rec_infer.tar) / [训练模型](https://paddleocr.bj.bcebos.com/dygraph_v2.0/multilingual/en_number_mobile_v2.0_rec_train.tar) |
<a name="多语言识别模型"></a> <a name="多语言识别模型"></a>
...@@ -64,11 +65,5 @@ PaddleOCR提供的可下载模型包括`推理模型`、`训练模型`、`预训 ...@@ -64,11 +65,5 @@ PaddleOCR提供的可下载模型包括`推理模型`、`训练模型`、`预训
|模型名称|模型简介|配置文件|推理模型大小|下载地址| |模型名称|模型简介|配置文件|推理模型大小|下载地址|
| --- | --- | --- | --- | --- | | --- | --- | --- | --- | --- |
|ch_ppocr_mobile_slim_v2.0_cls|slim量化版模型|[cls_mv3.yml](../../configs/cls/cls_mv3.yml)| |[推理模型 (coming soon)](link) / [训练模型](link) / [slim模型](link) | |ch_ppocr_mobile_slim_v2.0_cls|slim量化版模型|[cls_mv3.yml](../../configs/cls/cls_mv3.yml)| |推理模型 (coming soon) / 训练模型 / slim模型 |
|ch_ppocr_mobile_v2.0_cls|原始模型|[cls_mv3.yml](../../configs/cls/cls_mv3.yml)|1.38M|[推理模型](https://paddleocr.bj.bcebos.com/dygraph_v2.0/ch/ch_ppocr_mobile_v2.0_cls_infer.tar) / [训练模型](https://paddleocr.bj.bcebos.com/dygraph_v2.0/ch/ch_ppocr_mobile_v2.0_cls_train.tar) | |ch_ppocr_mobile_v2.0_cls|原始模型|[cls_mv3.yml](../../configs/cls/cls_mv3.yml)|1.38M|[推理模型](https://paddleocr.bj.bcebos.com/dygraph_v2.0/ch/ch_ppocr_mobile_v2.0_cls_infer.tar) / [训练模型](https://paddleocr.bj.bcebos.com/dygraph_v2.0/ch/ch_ppocr_mobile_v2.0_cls_train.tar) |
## OCR模型列表（V1.1，2020年9月22日更新）
[1.1系列模型地址](https://github.com/PaddlePaddle/PaddleOCR/blob/dygraph/doc/doc_ch/models_list.md)
doc/doc_ch/recognition.md
View file @ 78d51971
...@@ -7,7 +7,7 @@ ...@@ -7,7 +7,7 @@
- [字典](#字典) - [字典](#字典)
- [支持空格](#支持空格) - [支持空格](#支持空格)
- [二、启动训练](#文本检测模型推理) - [二、启动训练](#启动训练)
- [1. 数据增强](#数据增强) - [1. 数据增强](#数据增强)
- [2. 训练](#训练) - [2. 训练](#训练)
- [3. 小语种](#小语种) - [3. 小语种](#小语种)
...@@ -167,7 +167,7 @@ tar -xf rec_mv3_none_bilstm_ctc_v2.0_train.tar && rm -rf rec_mv3_none_bilstm_ctc ...@@ -167,7 +167,7 @@ tar -xf rec_mv3_none_bilstm_ctc_v2.0_train.tar && rm -rf rec_mv3_none_bilstm_ctc
``` ```
# GPU训练 支持单卡，多卡训练，通过--gpus参数指定卡号 # GPU训练 支持单卡，多卡训练，通过--gpus参数指定卡号
# 训练icdar15英文数据 并将训练日志保存为 tain_rec.log # 训练icdar15英文数据 训练日志会自动保存为 "{save_model_dir}" 下的train.log
python3 -m paddle.distributed.launch --gpus '0,1,2,3' tools/train.py -c configs/rec/rec_icdar15_train.yml python3 -m paddle.distributed.launch --gpus '0,1,2,3' tools/train.py -c configs/rec/rec_icdar15_train.yml
``` ```
<a name="数据增强"></a> <a name="数据增强"></a>
...@@ -200,11 +200,8 @@ PaddleOCR支持训练和评估交替进行, 可以在 `configs/rec/rec_icdar15_t ...@@ -200,11 +200,8 @@ PaddleOCR支持训练和评估交替进行, 可以在 `configs/rec/rec_icdar15_t
| rec_icdar15_train.yml | CRNN | Mobilenet_v3 large 0.5 | None | BiLSTM | ctc | | rec_icdar15_train.yml | CRNN | Mobilenet_v3 large 0.5 | None | BiLSTM | ctc |
| rec_mv3_none_bilstm_ctc.yml | CRNN | Mobilenet_v3 large 0.5 | None | BiLSTM | ctc | | rec_mv3_none_bilstm_ctc.yml | CRNN | Mobilenet_v3 large 0.5 | None | BiLSTM | ctc |
| rec_mv3_none_none_ctc.yml | Rosetta | Mobilenet_v3 large 0.5 | None | None | ctc | | rec_mv3_none_none_ctc.yml | Rosetta | Mobilenet_v3 large 0.5 | None | None | ctc |
| rec_mv3_tps_bilstm_ctc.yml | STARNet | Mobilenet_v3 large 0.5 | tps | BiLSTM | ctc |
| rec_mv3_tps_bilstm_attn.yml | RARE | Mobilenet_v3 large 0.5 | tps | BiLSTM | attention |
| rec_r34_vd_none_bilstm_ctc.yml | CRNN | Resnet34_vd | None | BiLSTM | ctc | | rec_r34_vd_none_bilstm_ctc.yml | CRNN | Resnet34_vd | None | BiLSTM | ctc |
| rec_r34_vd_none_none_ctc.yml | Rosetta | Resnet34_vd | None | None | ctc | | rec_r34_vd_none_none_ctc.yml | Rosetta | Resnet34_vd | None | None | ctc |
| rec_r34_vd_tps_bilstm_ctc.yml | STARNet | Resnet34_vd | tps | BiLSTM | ctc |
训练中文数据，推荐使用[rec_chinese_lite_train_v2.0.yml](../../configs/rec/ch_ppocr_v2.0/rec_chinese_lite_train_v2.0.yml)，如您希望尝试其他算法在中文数据集上的效果，请参考下列说明修改配置文件： 训练中文数据，推荐使用[rec_chinese_lite_train_v2.0.yml](../../configs/rec/ch_ppocr_v2.0/rec_chinese_lite_train_v2.0.yml)，如您希望尝试其他算法在中文数据集上的效果，请参考下列说明修改配置文件：
...@@ -356,8 +353,7 @@ python3 tools/infer_rec.py -c configs/rec/rec_icdar15_train.yml -o Global.checkp ...@@ -356,8 +353,7 @@ python3 tools/infer_rec.py -c configs/rec/rec_icdar15_train.yml -o Global.checkp
``` ```
infer_img: doc/imgs_words/en/word_1.png infer_img: doc/imgs_words/en/word_1.png
index: [19 24 18 23 29] result: ('joint', 0.9998967)
word : joint
``` ```
预测使用的配置文件必须与训练一致，如您通过 `python3 tools/train.py -c configs/rec/ch_ppocr_v2.0/rec_chinese_lite_train_v2.0.yml` 完成了中文模型的训练， 预测使用的配置文件必须与训练一致，如您通过 `python3 tools/train.py -c configs/rec/ch_ppocr_v2.0/rec_chinese_lite_train_v2.0.yml` 完成了中文模型的训练，
...@@ -376,6 +372,5 @@ python3 tools/infer_rec.py -c configs/rec/ch_ppocr_v2.0/rec_chinese_lite_train_v ...@@ -376,6 +372,5 @@ python3 tools/infer_rec.py -c configs/rec/ch_ppocr_v2.0/rec_chinese_lite_train_v
``` ```
infer_img: doc/imgs_words/ch/word_1.jpg infer_img: doc/imgs_words/ch/word_1.jpg
index: [2092 177 312 2503] result: ('韩国小馆', 0.997218)
word : 韩国小馆
``` ```
doc/doc_ch/update.md
View file @ 78d51971
# 更新 # 更新
- 2020.12.15 更新数据合成工具[Style-Text](../../StyleText/README_ch.md)，可以批量合成大量与目标场景类似的图像，在多个场景验证，效果明显提升。
- 2020.12.07 [FAQ](../../doc/doc_ch/FAQ.md)新增5个高频问题，总数124个，并且计划以后每周一都会更新，欢迎大家持续关注。
- 2020.11.25 更新半自动标注工具[PPOCRLabel](../../PPOCRLabel/README_ch.md)，辅助开发者高效完成标注任务，输出格式与PP-OCR训练任务完美衔接。
- 2020.9.22 更新PP-OCR技术文章，https://arxiv.org/abs/2009.09941 - 2020.9.22 更新PP-OCR技术文章，https://arxiv.org/abs/2009.09941
- 2020.9.19 更新超轻量压缩ppocr_mobile_slim系列模型，整体模型3.5M(详见[PP-OCR Pipline](../../README_ch.md#PP-OCR))，适合在移动端部署使用。[模型下载](../../README_ch.md#模型下载) - 2020.9.19 更新超轻量压缩ppocr_mobile_slim系列模型，整体模型3.5M(详见PP-OCR Pipline)，适合在移动端部署使用。
- 2020.9.17 更新超轻量ppocr_mobile系列和通用ppocr_server系列中英文ocr模型，媲美商业效果。[模型下载](../../README_ch.md#模型下载) - 2020.9.17 更新超轻量ppocr_mobile系列和通用ppocr_server系列中英文ocr模型，媲美商业效果。
- 2020.9.17 更新[英文识别模型](./models_list.md#english-recognition-model)和[多语种识别模型](./models_list.md#english-recognition-model)，已支持`德语、法语、日语、韩语`，更多语种识别模型将持续更新。 - 2020.9.17 更新[英文识别模型](./models_list.md#english-recognition-model)和[多语种识别模型](./models_list.md#english-recognition-model)，已支持`德语、法语、日语、韩语`，更多语种识别模型将持续更新。
- 2020.8.26 更新OCR相关的84个常见问题及解答，具体参考[FAQ](./FAQ.md) - 2020.8.26 更新OCR相关的84个常见问题及解答，具体参考[FAQ](./FAQ.md)
- 2020.8.24 支持通过whl包安装使用PaddleOCR，具体参考[Paddleocr Package使用说明](https://github.com/PaddlePaddle/PaddleOCR/blob/develop/doc/doc_ch/whl.md) - 2020.8.24 支持通过whl包安装使用PaddleOCR，具体参考[Paddleocr Package使用说明](https://github.com/PaddlePaddle/PaddleOCR/blob/develop/doc/doc_ch/whl.md)
......
doc/doc_ch/visualization.md
View file @ 78d51971
# 效果展示 # 效果展示
<a name="通用ppocr_server_1.1效果展示"></a> <a name="超轻量ppocr_server_2.0效果展示"></a>
## 通用ppocr_server_1.1效果展示 ## 通用ppocr_server_2.0 效果展示
<div align="center"> <div align="center">
<img src="../imgs_results/1101.jpg" width="800"> <img src="../imgs_results/ch_ppocr_mobile_v2.0/00006737.jpg" width="800">
<img src="../imgs_results/1102.jpg" width="800"> <img src="../imgs_results/ch_ppocr_mobile_v2.0/00009282.jpg" width="800">
<img src="../imgs_results/1103.jpg" width="800"> <img src="../imgs_results/ch_ppocr_mobile_v2.0/00015504.jpg" width="800">
<img src="../imgs_results/1104.jpg" width="800"> <img src="../imgs_results/ch_ppocr_mobile_v2.0/00018069.jpg" width="800">
<img src="../imgs_results/1105.jpg" width="800"> <img src="../imgs_results/ch_ppocr_mobile_v2.0/00056221.jpg" width="800">
<img src="../imgs_results/1106.jpg" width="800"> <img src="../imgs_results/ch_ppocr_mobile_v2.0/00057937.jpg" width="800">
<img src="../imgs_results/ch_ppocr_mobile_v2.0/00059985.jpg" width="800">
<img src="../imgs_results/ch_ppocr_mobile_v2.0/00111002.jpg" width="800">
<img src="../imgs_results/ch_ppocr_mobile_v2.0/00077949.jpg" width="800">
<img src="../imgs_results/ch_ppocr_mobile_v2.0/00207393.jpg" width="800">
</div> </div>
<a name="英文识别模型效果展示"></a> <a name="英文识别模型效果展示"></a>
## 英文识别模型效果展示 ## 英文识别模型效果展示
<div align="center"> <div align="center">
<img src="../imgs_results/img_12.jpg" width="800"> <img src="../imgs_results/ch_ppocr_mobile_v2.0/img_12.jpg" width="800">
</div> </div>
<a name="多语言识别模型效果展示"></a> <a name="多语言识别模型效果展示"></a>
## 多语言识别模型效果展示 ## 多语言识别模型效果展示
<div align="center"> <div align="center">
<img src="../imgs_results/1110.jpg" width="800"> <img src="../imgs_results/french_0.jpg" width="800">
<img src="../imgs_results/1112.jpg" width="800"> <img src="../imgs_results/korean.jpg" width="800">
</div>
<a name="超轻量ppocr_mobile_1.0效果展示"></a>
## 超轻量ppocr_mobile_1.0效果展示
<div align="center">
<img src="../imgs_results/1.jpg" width="800">
<img src="../imgs_results/7.jpg" width="800">
<img src="../imgs_results/6.jpg" width="800">
<img src="../imgs_results/16.png" width="800">
</div>
<a name="通用ppocr_server_1.0效果展示"></a>
## 通用ppocr_server_1.0效果展示
<div align="center">
<img src="../imgs_results/chinese_db_crnn_server/11.jpg" width="800">
<img src="../imgs_results/chinese_db_crnn_server/2.jpg" width="800">
<img src="../imgs_results/chinese_db_crnn_server/8.jpg" width="800">
</div> </div>
doc/doc_ch/whl.md
View file @ 78d51971
...@@ -6,7 +6,7 @@ ...@@ -6,7 +6,7 @@
pip安装 pip安装
```bash ```bash
pip install paddleocr pip install "paddleocr>=2.0.1" # 推荐使用2.0.1+版本
``` ```
本地构建并安装 本地构建并安装
...@@ -166,7 +166,7 @@ paddleocr -h ...@@ -166,7 +166,7 @@ paddleocr -h
* 检测+分类+识别全流程 * 检测+分类+识别全流程
```bash ```bash
paddleocr --image_dir PaddleOCR/doc/imgs/11.jpg --use_angle_cls true --cls true paddleocr --image_dir PaddleOCR/doc/imgs/11.jpg --use_angle_cls true
``` ```
结果是一个list，每个item包含了文本框，文字和识别置信度 结果是一个list，每个item包含了文本框，文字和识别置信度
```bash ```bash
...@@ -190,7 +190,7 @@ paddleocr --image_dir PaddleOCR/doc/imgs/11.jpg ...@@ -190,7 +190,7 @@ paddleocr --image_dir PaddleOCR/doc/imgs/11.jpg
* 分类+识别 * 分类+识别
```bash ```bash
paddleocr --image_dir PaddleOCR/doc/imgs_words/ch/word_1.jpg --use_angle_cls true --cls true --det false paddleocr --image_dir PaddleOCR/doc/imgs_words/ch/word_1.jpg --use_angle_cls true --det false
``` ```
结果是一个list，每个item只包含识别结果和识别置信度 结果是一个list，每个item只包含识别结果和识别置信度
...@@ -222,7 +222,7 @@ paddleocr --image_dir PaddleOCR/doc/imgs_words/ch/word_1.jpg --det false ...@@ -222,7 +222,7 @@ paddleocr --image_dir PaddleOCR/doc/imgs_words/ch/word_1.jpg --det false
* 单独执行分类 * 单独执行分类
```bash ```bash
paddleocr --image_dir PaddleOCR/doc/imgs_words/ch/word_1.jpg --use_angle_cls true --cls true --det false --rec false paddleocr --image_dir PaddleOCR/doc/imgs_words/ch/word_1.jpg --use_angle_cls true --det false --rec false
``` ```
结果是一个list，每个item只包含分类结果和分类置信度 结果是一个list，每个item只包含分类结果和分类置信度
...@@ -258,7 +258,7 @@ im_show.save('result.jpg') ...@@ -258,7 +258,7 @@ im_show.save('result.jpg')
### 通过命令行使用 ### 通过命令行使用
```bash ```bash
paddleocr --image_dir PaddleOCR/doc/imgs/11.jpg --det_model_dir {your_det_model_dir} --rec_model_dir {your_rec_model_dir} --rec_char_dict_path {your_rec_char_dict_path} --cls_model_dir {your_cls_model_dir} --use_angle_cls true --cls true paddleocr --image_dir PaddleOCR/doc/imgs/11.jpg --det_model_dir {your_det_model_dir} --rec_model_dir {your_rec_model_dir} --rec_char_dict_path {your_rec_char_dict_path} --cls_model_dir {your_cls_model_dir} --use_angle_cls true
``` ```
### 使用网络图片或者numpy数组作为输入 ### 使用网络图片或者numpy数组作为输入
......
doc/doc_en/algorithm_overview_en.md 100644 → 100755
View file @ 78d51971
<a name="Algorithm_introduction"></a> <a name="Algorithm_introduction"></a>
## Algorithm introduction ## Algorithm introduction
This tutorial lists the text detection algorithms and text recognition algorithms supported by PaddleOCR, as well as the models and metrics of each algorithm on **English public datasets**. It is mainly used for algorithm introduction and algorithm performance comparison. For more models on other datasets including Chinese, please refer to [PP-OCR v1.1 models list](./models_list_en.md). This tutorial lists the text detection algorithms and text recognition algorithms supported by PaddleOCR, as well as the models and metrics of each algorithm on **English public datasets**. It is mainly used for algorithm introduction and algorithm performance comparison. For more models on other datasets including Chinese, please refer to [PP-OCR v2.0 models list](./models_list_en.md).
- [1. Text Detection Algorithm](#TEXTDETECTIONALGORITHM) - [1. Text Detection Algorithm](#TEXTDETECTIONALGORITHM)
...@@ -13,27 +13,27 @@ This tutorial lists the text detection algorithms and text recognition algorithm ...@@ -13,27 +13,27 @@ This tutorial lists the text detection algorithms and text recognition algorithm
PaddleOCR open source text detection algorithms list: PaddleOCR open source text detection algorithms list:
- [x] EAST([paper](https://arxiv.org/abs/1704.03155)) - [x] EAST([paper](https://arxiv.org/abs/1704.03155))
- [x] DB([paper](https://arxiv.org/abs/1911.08947)) - [x] DB([paper](https://arxiv.org/abs/1911.08947))
- [x] SAST([paper](https://arxiv.org/abs/1908.05498))(Baidu Self-Research) - [x] SAST([paper](https://arxiv.org/abs/1908.05498) )(Baidu Self-Research)
On the ICDAR2015 dataset, the text detection result is as follows: On the ICDAR2015 dataset, the text detection result is as follows:
|Model|Backbone|precision|recall|Hmean|Download link| |Model|Backbone|precision|recall|Hmean|Download link|
|-|-|-|-|-|-| | --- | --- | --- | --- | --- | --- |
|EAST|ResNet50_vd|88.18%|85.51%|86.82%|[Download link](link)| |EAST|ResNet50_vd|88.76%|81.36%|84.90%|[Download link](https://paddleocr.bj.bcebos.com/dygraph_v2.0/en/det_r50_vd_east_v2.0_train.tar)|
|EAST|MobileNetV3|81.67%|79.83%|80.74%|[Download link](link)| |EAST|MobileNetV3|78.24%|79.15%|78.69%|[Download link](https://paddleocr.bj.bcebos.com/dygraph_v2.0/en/det_mv3_east_v2.0_train.tar)|
|DB|ResNet50_vd|83.79%|80.65%|82.19%|[Download link](link)| |DB|ResNet50_vd|86.41%|78.72%|82.38%|[Download link](https://paddleocr.bj.bcebos.com/dygraph_v2.0/en/det_r50_vd_db_v2.0_train.tar)|
|DB|MobileNetV3|75.92%|73.18%|74.53%|[Download link](link)| |DB|MobileNetV3|77.29%|73.08%|75.12%|[Download link](https://paddleocr.bj.bcebos.com/dygraph_v2.0/en/det_mv3_db_v2.0_train.tar)|
|SAST|ResNet50_vd|92.18%|82.96%|87.33%|[Download link](link)| |SAST|ResNet50_vd|91.83%|81.80%|86.52%|[Download link](https://paddleocr.bj.bcebos.com/dygraph_v2.0/en/det_r50_vd_sast_icdar15_v2.0_train.tar)|
On Total-Text dataset, the text detection result is as follows: On Total-Text dataset, the text detection result is as follows:
|Model|Backbone|precision|recall|Hmean|Download link| |Model|Backbone|precision|recall|Hmean|Download link|
|-|-|-|-|-|-| | --- | --- | --- | --- | --- | --- |
|SAST|ResNet50_vd|88.74%|79.80%|84.03%|[Download link](link)| |SAST|ResNet50_vd|89.05%|76.80%|82.47%|[Download link](https://paddleocr.bj.bcebos.com/dygraph_v2.0/en/det_r50_vd_sast_totaltext_v2.0_train.tar)|
**Note：** Additional data, like icdar2013, icdar2017, COCO-Text, ArT, was added to the model training of SAST. Download English public dataset in organized format used by PaddleOCR from [Baidu Drive](https://pan.baidu.com/s/12cPnZcVuV1zn5DOd4mqjVw) (download code: 2bpi). **Note：** Additional data, like icdar2013, icdar2017, COCO-Text, ArT, was added to the model training of SAST. Download English public dataset in organized format used by PaddleOCR from [Baidu Drive](https://pan.baidu.com/s/12cPnZcVuV1zn5DOd4mqjVw) (download code: 2bpi).
For the training guide and use of PaddleOCR text detection algorithms, please refer to the document [Text detection model training/evaluation/prediction](./doc/doc_en/detection_en.md) For the training guide and use of PaddleOCR text detection algorithms, please refer to the document [Text detection model training/evaluation/prediction](./detection_en.md)
<a name="TEXTRECOGNITIONALGORITHM"></a> <a name="TEXTRECOGNITIONALGORITHM"></a>
### 2. Text Recognition Algorithm ### 2. Text Recognition Algorithm
...@@ -41,20 +41,17 @@ For the training guide and use of PaddleOCR text detection algorithms, please re ...@@ -41,20 +41,17 @@ For the training guide and use of PaddleOCR text detection algorithms, please re
PaddleOCR open-source text recognition algorithms list: PaddleOCR open-source text recognition algorithms list:
- [x] CRNN([paper](https://arxiv.org/abs/1507.05717)) - [x] CRNN([paper](https://arxiv.org/abs/1507.05717))
- [x] Rosetta([paper](https://arxiv.org/abs/1910.05085)) - [x] Rosetta([paper](https://arxiv.org/abs/1910.05085))
- [x] STAR-Net([paper](http://www.bmva.org/bmvc/2016/papers/paper043/index.html)) - [ ] STAR-Net([paper](http://www.bmva.org/bmvc/2016/papers/paper043/index.html)) coming soon
- [ ] RARE([paper](https://arxiv.org/abs/1603.03915v1)) coming soon - [ ] RARE([paper](https://arxiv.org/abs/1603.03915v1)) coming soon
- [ ] SRN([paper](https://arxiv.org/abs/2003.12294))(Baidu Self-Research) coming soon - [ ] SRN([paper](https://arxiv.org/abs/2003.12294) )(Baidu Self-Research) coming soon
Refer to [DTRB](https://arxiv.org/abs/1904.01906), the training and evaluation result of these above text recognition (using MJSynth and SynthText for training, evaluate on IIIT, SVT, IC03, IC13, IC15, SVTP, CUTE) is as follow: Refer to [DTRB](https://arxiv.org/abs/1904.01906), the training and evaluation result of these above text recognition (using MJSynth and SynthText for training, evaluate on IIIT, SVT, IC03, IC13, IC15, SVTP, CUTE) is as follow:
|Model|Backbone|Avg Accuracy|Module combination|Download link| |Model|Backbone|Avg Accuracy|Module combination|Download link|
|-|-|-|-|-| |-|-|-|-|-|
|Rosetta|Resnet34_vd|80.24%|rec_r34_vd_none_none_ctc|[Download link](https://paddleocr.bj.bcebos.com/rec_r34_vd_none_none_ctc.tar)| |Rosetta|Resnet34_vd|80.9%|rec_r34_vd_none_none_ctc|[Download link](https://paddleocr.bj.bcebos.com/dygraph_v2.0/en/rec_r34_vd_none_none_ctc_v2.0_train.tar)|
|Rosetta|MobileNetV3|78.16%|rec_mv3_none_none_ctc|[Download link](https://paddleocr.bj.bcebos.com/rec_mv3_none_none_ctc.tar)| |Rosetta|MobileNetV3|78.05%|rec_mv3_none_none_ctc|[Download link](https://paddleocr.bj.bcebos.com/dygraph_v2.0/en/rec_mv3_none_none_ctc_v2.0_train.tar)|
|CRNN|Resnet34_vd|82.20%|rec_r34_vd_none_bilstm_ctc|[Download link](https://paddleocr.bj.bcebos.com/rec_r34_vd_none_bilstm_ctc.tar)| |CRNN|Resnet34_vd|82.76%|rec_r34_vd_none_bilstm_ctc|[Download link](https://paddleocr.bj.bcebos.com/dygraph_v2.0/en/rec_r34_vd_none_bilstm_ctc_v2.0_train.tar)|
|CRNN|MobileNetV3|79.37%|rec_mv3_none_bilstm_ctc|[Download link](https://paddleocr.bj.bcebos.com/rec_mv3_none_bilstm_ctc.tar)| |CRNN|MobileNetV3|79.97%|rec_mv3_none_bilstm_ctc|[Download link](https://paddleocr.bj.bcebos.com/dygraph_v2.0/en/rec_mv3_none_bilstm_ctc_v2.0_train.tar)|
|STAR-Net|Resnet34_vd|83.93%|rec_r34_vd_tps_bilstm_ctc|[Download link](https://paddleocr.bj.bcebos.com/rec_r34_vd_tps_bilstm_ctc.tar)|
|STAR-Net|MobileNetV3|81.56%|rec_mv3_tps_bilstm_ctc|[Download link](https://paddleocr.bj.bcebos.com/rec_mv3_tps_bilstm_ctc.tar)|
Please refer to the document for training guide and use of PaddleOCR text recognition algorithms [Text recognition model training/evaluation/prediction](./recognition_en.md)
Please refer to the document for training guide and use of PaddleOCR text recognition algorithms [Text recognition model training/evaluation/prediction](./doc/doc_en/recognition_en.md)
doc/doc_en/angle_class_en.md
View file @ 78d51971
...@@ -65,9 +65,9 @@ Start training: ...@@ -65,9 +65,9 @@ Start training:
``` ```
# Set PYTHONPATH path # Set PYTHONPATH path
export PYTHONPATH=$PYTHONPATH:. export PYTHONPATH=$PYTHONPATH:.
# GPU training Support single card and multi-card training, specify the card number through selected_gpus # GPU training Support single card and multi-card training, specify the card number through --gpus. If your paddle version is less than 2.0rc1, please use '--selected_gpus'
# Start training, the following command has been written into the train.sh file, just modify the configuration file path in the file # Start training, the following command has been written into the train.sh file, just modify the configuration file path in the file
python3 -m paddle.distributed.launch --selected_gpus '0,1,2,3,4,5,6,7' tools/train.py -c configs/cls/cls_mv3.yml python3 -m paddle.distributed.launch --gpus '0,1,2,3,4,5,6,7' tools/train.py -c configs/cls/cls_mv3.yml
``` ```
- Data Augmentation - Data Augmentation
...@@ -77,7 +77,7 @@ PaddleOCR provides a variety of data augmentation methods. If you want to add di ...@@ -77,7 +77,7 @@ PaddleOCR provides a variety of data augmentation methods. If you want to add di
The default perturbation methods are: cvtColor, blur, jitter, Gasuss noise, random crop, perspective, color reverse, RandAugment. The default perturbation methods are: cvtColor, blur, jitter, Gasuss noise, random crop, perspective, color reverse, RandAugment.
Except for RandAugment, each disturbance method is selected with a 50% probability during the training process. For specific code implementation, please refer to: Except for RandAugment, each disturbance method is selected with a 50% probability during the training process. For specific code implementation, please refer to:
[rec_img_aug.py](../../ppocr/data/imaug/rec_img_aug.py) [rec_img_aug.py](../../ppocr/data/imaug/rec_img_aug.py)
[randaugment.py](../../ppocr/data/imaug/randaugment.py) [randaugment.py](../../ppocr/data/imaug/randaugment.py)
......
doc/doc_en/benchmark_en.md 100644 → 100755
View file @ 78d51971
File mode changed from 100644 to 100755
doc/doc_en/detection_en.md
View file @ 78d51971
...@@ -76,8 +76,10 @@ You can also use `-o` to change the training parameters without modifying the ym ...@@ -76,8 +76,10 @@ You can also use `-o` to change the training parameters without modifying the ym
python3 tools/train.py -c configs/det/det_mv3_db.yml -o Optimizer.base_lr=0.0001 python3 tools/train.py -c configs/det/det_mv3_db.yml -o Optimizer.base_lr=0.0001
# multi-GPU training # multi-GPU training
# Set the GPU ID used by the '--select_gpus' parameter; # Set the GPU ID used by the '--gpus' parameter; If your paddle version is less than 2.0rc1, please use '--selected_gpus'
python3 -m paddle.distributed.launch --selected_gpus '0,1,2,3' tools/train.py -c configs/det/det_mv3_db.yml -o Optimizer.base_lr=0.0001 python3 -m paddle.distributed.launch --gpus '0,1,2,3' tools/train.py -c configs/det/det_mv3_db.yml -o Optimizer.base_lr=0.0001
``` ```
#### load trained model and continue training #### load trained model and continue training
...@@ -99,15 +101,11 @@ Run the following code to calculate the evaluation indicators. The result will b ...@@ -99,15 +101,11 @@ Run the following code to calculate the evaluation indicators. The result will b
When evaluating, set post-processing parameters `box_thresh=0.6`, `unclip_ratio=1.5`. If you use different datasets, different models for training, these two parameters should be adjusted for better result. When evaluating, set post-processing parameters `box_thresh=0.6`, `unclip_ratio=1.5`. If you use different datasets, different models for training, these two parameters should be adjusted for better result.
The model parameters during training are saved in the `Global.save_model_dir` directory by default. When evaluating indicators, you need to set `Global.checkpoints` to point to the saved parameter file.
```shell ```shell
python3 tools/eval.py -c configs/det/det_mv3_db.yml -o Global.checkpoints="{path/to/weights}/best_accuracy" PostProcess.box_thresh=0.6 PostProcess.unclip_ratio=1.5 python3 tools/eval.py -c configs/det/det_mv3_db.yml -o Global.checkpoints="{path/to/weights}/best_accuracy" PostProcess.box_thresh=0.6 PostProcess.unclip_ratio=1.5
``` ```
The model parameters during training are saved in the `Global.save_model_dir` directory by default. When evaluating indicators, you need to set `Global.checkpoints` to point to the saved parameter file.
Such as:
```shell
python3 tools/eval.py -c configs/det/det_mv3_db.yml -o Global.checkpoints="./output/det_db/best_accuracy" PostProcess.box_thresh=0.6 PostProcess.unclip_ratio=1.5
```
* Note: `box_thresh` and `unclip_ratio` are parameters required for DB post-processing, and not need to be set when evaluating the EAST model. * Note: `box_thresh` and `unclip_ratio` are parameters required for DB post-processing, and not need to be set when evaluating the EAST model.
......
Markdown is supported
0% or .
You are about to add 0 people to the discussion. Proceed with caution.
Finish editing this message first!
Please register or to comment