添加onnx转rknn脚本

010使用PaddleOCR_rknn.py

@@ -307,6 +307,8 @@ class RK3588OCR:
         # 推理
         det_output = self.det_rknn.inference(inputs=[det_img], data_format="nchw")[0]
         
+        print("*"*100, np.max(det_output))
+        
         # 后处理
         mask = det_output[0, 0, :, :]
         threshold = 0.3
@@ -407,7 +409,7 @@ class RK3588OCR:
             # 识别文本
             text, conf = self.recognize_text(crop_img)
             
-            if conf > 0.4:  # 置信度过滤
+            if conf > 0.1:  # 置信度过滤
                 ocr_results.append({
                     'text': text,
                     'confidence': conf,

011onnx转rknn.py

@@ -0,0 +1,58 @@
+from rknn.api import RKNN
+import cv2
+import numpy as np
+
+
+# 初始化 RKNN
+rknn = RKNN()
+
+# 配置参数（关键！）
+rknn.config(
+    target_platform="rk3588",          # 根据实际芯片型号修改
+    mean_values=[[0, 0, 0]],       # YOLOv8 输入为 0-255，无需归一化
+    std_values=[[255, 255, 255]],  # 输入数据除以 255（即 0-1 范围）
+    # quant_img_RGB2BGR=True,
+    optimization_level=3,          # 最高优化级别
+)
+
+# 加载 ONNX
+ret = rknn.load_onnx(model="/home/orangepi/Desktop/kangda_robotic_dog/yolov8_20250820.onnx")
+assert ret == 0, "加载 ONNX 失败！"
+
+# 转换模型
+ret = rknn.build(
+    do_quantization=False,          # 启用量化
+    # dataset="dataset.txt",         # 校准数据路径
+)
+assert ret == 0, "转换 RKNN 失败！"
+
+# 导出 RKNN
+ret = rknn.export_rknn("/home/orangepi/Desktop/kangda_robotic_dog/yolov8_20250820.rknn")
+assert ret == 0, "导出 RKNN 失败！"
+
+
+# Set inputs
+img = cv2.imread('/home/orangepi/Desktop/kangda_robotic_dog/微信图片_20250827165826.jpg')
+img = cv2.cvtColor(img, cv2.COLOR_BGR2RGB)
+img.resize((3, 640, 640))
+img = np.expand_dims(img, 0)
+
+# Init runtime environment
+print('--> Init runtime environment')
+ret = rknn.init_runtime()
+if ret != 0:
+    print('Init runtime environment failed!')
+    exit(ret)
+print('done')
+
+# Inference
+print('--> Running model')
+outputs = rknn.inference(inputs=[img], data_format=['nchw'])
+np.save('./tflite_mobilenet_v1_0.npy', outputs[0])
+print(len(outputs))
+print('done')
+
+rknn.release()
+
+
+

机器狗后台服务/app/util/baiduOCR.py

@@ -134,7 +134,7 @@ class BaiduOCRONNX:
         
         return dt_boxes
     
-    def boxes_from_bitmap(self, pred, bitmap, dest_width, dest_height, max_candidates=1000, box_thresh=0.6):
+    def boxes_from_bitmap(self, pred, bitmap, dest_width, dest_height, max_candidates=1000, box_thresh=0.4):
         """
         从位图中提取文本框
         """
@@ -312,6 +312,8 @@ class BaiduOCRONNX:
         # 将坐标转换回原图空间
         if len(boxes) > 0:
             boxes = self.post_process_det(boxes, ratio, padding_info, (ori_h, ori_w))
+            
+        print("lenboxes", len(boxes))
         
         return boxes, scores
     
@@ -371,6 +373,9 @@ class BaiduOCRONNX:
         if dt_boxes is None or len(dt_boxes) == 0:
             return []
         
+        
+        # visualize_results(image_path, {"box":dt_boxes})
+        
         # 2. 文本识别
         ocr_results = []
         
@@ -404,22 +409,22 @@ def main():
     # 初始化OCR
     # ocr = PaddleOCRONNX('/home/admin-root/haotian/康达瑞贝斯机器狗/det_shape.onnx', '/home/admin-root/haotian/康达瑞贝斯机器狗/rec_shape.onnx')
     
-    ocr = BaiduOCRONNX('/home/admin-root/haotian/康达瑞贝斯机器狗/det_shape_20250814.onnx', '/home/admin-root/haotian/康达瑞贝斯机器狗/rec_shape_20250815.onnx')
+    ocr = BaiduOCRONNX('/home/orangepi/Desktop/kangda_robotic_dog/det_mobile_14_shape.onnx', '/home/orangepi/Desktop/kangda_robotic_dog/rec_mobile_14_shape.onnx')
     
     # 执行OCR
-    image_path = '/home/admin-root/haotian/康达瑞贝斯机器狗/data_image/001读表图片/3aee64cc1f90d93a5a45979f7b17cb4b_frame_001460.jpg'
+    image_path = '/home/orangepi/Desktop/kangda_robotic_dog/image_test/微信图片_20250827152914.jpg'
     results = ocr.ocr(image_path)
     
     # 打印结果
-    for result in results:
-        print(f"文本: {result['text']}")
-        print(f"置信度: {result['confidence']:.3f}")
-        print(f"检测得分: {result['score']:.3f}")
-        print(f"坐标: {result['box']}")
-        print("-" * 50)
+    # for result in results:
+    #     print(f"文本: {result['text']}")
+    #     print(f"置信度: {result['confidence']:.3f}")
+    #     print(f"检测得分: {result['score']:.3f}")
+    #     print(f"坐标: {result['box']}")
+    #     print("-" * 50)
     
     # 可视化结果
-    visualize_results(image_path, results)
+    # visualize_results(image_path, results)
 
 def visualize_results(image_path, results):
     """
@@ -442,5 +447,7 @@ def visualize_results(image_path, results):
     
     
 if __name__ == '__main__':
-    ocr = BaiduOCR()
-    print(ocr.ocr(""))
+    
+    main()
+    # ocr = BaiduOCR()
+    # print(ocr.ocr(""))

机器狗后台服务/test_baiduocr.py

@@ -0,0 +1,3 @@
+from app.util.baiduOCR import main
+
+main()

机器狗后台服务/test_base64_server.py

@@ -82,20 +82,22 @@ if __name__ == "__main__":
     # 测试图片路径，请根据实际情况修改
     # test_image_path = "/home/admin-root/haotian/康达瑞贝斯机器狗/data_image/001读表图片/2c7cc83019e7388a7041101da92c9829_frame_000000.jpg"
     
-    # #---------------------------------------测试ocr-----------------------------------------
-    import time
-    test_image_path = "/home/orangepi/Desktop/kangda_robotic_dog/image_test/632e474452d560edd7004f745319ff00_frame_000730.jpg"
+    # # #---------------------------------------测试ocr-----------------------------------------
+    # import time
+    # # test_image_path = "/home/orangepi/Desktop/kangda_robotic_dog/image_test/632e474452d560edd7004f745319ff00_frame_000300.jpg"
     
-    api_url="http://10.0.0.81:12345/api/v1/ocr_from_base64"
-    start_time = time.time()
-    # 调用测试函数
-    test_ocr_api(test_image_path, api_url)
-    print("cost time: ", time.time()-start_time)
-    # #---------------------------------------测试ocrender-----------------------------------------
+    # test_image_path = "/home/orangepi/Desktop/kangda_robotic_dog/投屏图片/微信图片_202508271731493.jpg"
+    
+    # api_url="http://10.0.0.81:12345/api/v1/ocr_from_base64"
+    # start_time = time.time()
+    # # 调用测试函数
+    # test_ocr_api(test_image_path, api_url)
+    # print("cost time: ", time.time()-start_time)
+    # # #---------------------------------------测试ocrender-----------------------------------------
     
 #    # -----------------------------------------测试yolov8 侵占消防区域检测-----------------------------------------
 #     # test_image_path = "/home/orangepi/Desktop/kangda_robotic_dog/image_test/1e4c75b76e531606e2adc491a8f09ae8_frame_000000.jpg"
-#     test_image_path = "/home/orangepi/Desktop/kangda_robotic_dog/image_test/1e4c75b76e531606e2adc491a8f09ae8_frame_000720.jpg"
+#     test_image_path = "/home/orangepi/Desktop/kangda_robotic_dog/投屏图片/微信图片_202508271731501.jpg"
 #     api_url = "http://10.0.0.81:12345/api/v1/detect_from_base64_0"
 #     test_detect(test_image_path, api_url)
 #     #-----------------------------------------测试yolov8 侵占消防区域检测 end-----------------------------------------
@@ -103,8 +105,8 @@ if __name__ == "__main__":
     
     #-----------------------------------------测试yolov8 灭火器检测-----------------------------------------
     # test_image_path = "/home/orangepi/Desktop/kangda_robotic_dog/image_test/ce81420a27cdaff14fe42f967eaa49a3_frame_001060.jpg"
-    # test_image_path = "/home/orangepi/Desktop/kangda_robotic_dog/image_test/1e4c75b76e531606e2adc491a8f09ae8_frame_000120.jpg"
-    # test_image_path = "/home/admin-root/haotian/康达瑞贝斯机器狗/YoloV8Obj/dataset_20250819/train/images/1e4c75b76e531606e2adc491a8f09ae8_frame_000120.jpg"
-    # api_url = "http://10.0.0.81:12345/api/v1/detect_from_base64_1"
-    # test_detect(test_image_path, api_url=api_url)
-    #-----------------------------------------测试yolov8 灭火器检测 end-----------------------------------------
+   # test_image_path = "/home/orangepi/Desktop/kangda_robotic_dog/image_test/1e4c75b76e531606e2adc491a8f09ae8_frame_000120.jpg"
+    test_image_path = "/home/orangepi/Desktop/kangda_robotic_dog/投屏图片/微信图片_202508271731496.jpg"
+    api_url = "http://10.0.0.81:12345/api/v1/detect_from_base64_1"
+    test_detect(test_image_path, api_url=api_url)
+   # -----------------------------------------测试yolov8 灭火器检测 end-----------------------------------------