提交crnn推理示例

Crnn_infer_migraphx.py

+import cv2
+import numpy as np
+import migraphx
+import time
+import argparse
+
+class Crnn:
+    def __init__(self, path):
+        # 解析推理模型
+        self.model = migraphx.parse_onnx(path)
+
+        # 获取模型的输入name
+        self.inputName = self.model.get_parameter_names()[0]
+
+        # 获取模型的输入尺寸
+        self.inputShape = self.model.get_parameter_shapes()[self.inputName].lens()
+
+        print("inputName:{0} \ninputShape:{1}".format(self.inputName, self.inputShape))
+
+        # 模型编译
+        self.model.compile(t=migraphx.get_target("gpu"), device_id=0)  # device_id: 设置GPU设备，默认为0号设备
+        print("Success to compile")
+
+    def infer(self, image):
+        inputImage = self.prepare_input(image)
+
+        # 执行推理
+        results = self.model.run({self.model.get_parameter_names()[0]: migraphx.argument(inputImage)})
+        # 获取第一个输出节点的数据,migraphx.argument类型
+        result=results[0]
+        result=np.array(result)
+        text = self.decode(result)
+        final_text = self.map_rule(text)
+        
+        return text, final_text
+
+    def prepare_input(self, image):
+        img_gray = cv2.cvtColor(image, cv2.COLOR_BGR2GRAY)
+        inputSize = (self.inputShape[3], self.inputShape[2])
+        blob = cv2.dnn.blobFromImage(img_gray, scalefactor=1 / 127.5, size=inputSize, mean=127.5)
+
+        return blob
+
+    def decode(self, scores):
+        alphabet = "-0123456789abcdefghijklmnopqrstuvwxyz"
+        text = ""
+
+        # 获取模型预测的文本序列
+        for i in range(scores.shape[0]):
+            c = np.argmax(scores[i][0])
+            text += alphabet[c]
+
+        return text
+
+    def map_rule(self, text):
+        char_list = []
+        for i in range(len(text)):
+            if i == 0:
+                if text[i] != '-':
+                    char_list.append(text[i])
+            else:
+                if text[i] != '-' and (not (text[i] == text[i - 1])):
+                    char_list.append(text[i])
+
+        return ''.join(char_list)
+
+if __name__ == '__main__':
+    parser = argparse.ArgumentParser()
+    parser.add_argument('--imgpath', type=str, default='./data/text.jpg', help="image path")
+    parser.add_argument('--modelpath', type=str, default='./weights/crnn.onnx', help="onnx filepath")
+    args = parser.parse_args()
+
+    crnn = Crnn(args.modelpath)
+
+    srcimg = cv2.imread(args.imgpath, 1)
+
+    # 执行推理
+    print("Start to inference")
+    start = time.time()
+    resultRaw, resultSim = crnn.infer(srcimg)
+    print('net forward time: {:.4f}'.format(time.time() - start))
+    print("============= Ocr Results =============")
+    print('%-20s => %-20s' % (resultRaw, resultSim))
+
+
+

README.md

-# CRNN_MIGraphX
+# CRNN
 
-This project constructs a CRNN character recognition inference example
+## 模型介绍
\ No newline at end of file
+
+CRNN是文本识别领域的一种经典算法，该算法的主要思想是认为文本识别需要对序列进行预测，所以采用了预测序列常用的RNN网络。算法通过CNN提取图片特征，然后采用RNN对序列进行预测，最终使用CTC方法得到最终结果。
+
+## 模型结构
+
+CRNN模型的主要结构包括基于CNN的图像特征提取模块以及基于双向LSTM的文字序列特征提取模块。
+
+## 推理
+
+### 环境配置
+
+在[光源](https://www.sourcefind.cn/#/image/dcu/custom)可拉取用于推理的docker镜像，CRNN 模型推理推荐的镜像如下：
+
+```
+docker pull image.sourcefind.cn:5000/dcu/admin/base/custom:ort_dcu_1.14.0_migraphx2.5.2_dtk22.10.1
+```
+
+在[光合开发者社区](https://cancon.hpccube.com:65024/4/main/)可下载MIGraphX安装包，python依赖安装：
+
+```
+pip install -r requirements.txt
+```
+
+### 运行示例
+
+RetinaFace模型的推理示例程序是Crnn_infer_migraphx.py，使用如下命令运行该推理示例：
+
+```
+python Crnn_infer_migraphx.py 
+```
+
+该示例输入样本图像为：
+
+<img src="./data/text.jpg" alt="Result"  />
+
+文本识别结果为：
+
+```
+a-----v--a-i-l-a-bb-l-e--- => available
+```
+
+## 历史版本
+
+​		https://developer.hpccube.com/codes/modelzoo/crnn_migraphx
+
+## 参考
+
+​		https://github.com/meijieru/crnn.pytorch

requirements.txt

+opencv-contrib-python
+numpy
+argparse
+time
\ No newline at end of file