文本检测模型微调完成

003测试paddleocr.py

@@ -2,7 +2,8 @@ from paddleocr import PaddleOCR
 
 ocr = PaddleOCR(
     # 文本检测模型地址
-    text_detection_model_dir = "/home/admin-root/haotian/康达瑞贝斯机器狗/ocr_model/PP-OCRv5_server_det",
+    # text_detection_model_dir = "/home/admin-root/haotian/康达瑞贝斯机器狗/ocr_model/PP-OCRv5_server_det",
+    text_detection_model_dir='/home/admin-root/haotian/康达瑞贝斯机器狗/PaddleOCR-3.1.0/output/PP-OCRv5_server_det_infer_20250814',
     # 文本识别模型地址
     text_recognition_model_dir = "/home/admin-root/haotian/康达瑞贝斯机器狗/ocr_model/PP-OCRv5_server_rec",
     use_doc_orientation_classify=False, 
@@ -16,7 +17,7 @@ ocr = PaddleOCR(
 #     use_doc_orientation_classify=False,
 #     use_doc_unwarping=False,
 #     use_textline_orientation=False) # 更换 PP-OCRv5_mobile 模型
-result = ocr.predict("data_image/001读表图片/2c7cc83019e7388a7041101da92c9829_frame_000000.jpg")
+result = ocr.predict("/home/admin-root/haotian/康达瑞贝斯机器狗/data_image/001读表图片/3aee64cc1f90d93a5a45979f7b17cb4b_frame_001460.jpg")
 for res in result:
     res.print()
     

PaddleOCR-3.1.0/configs/det/PP-OCRv5/PP-OCRv5_server_det.yml

@@ -71,7 +71,7 @@ Metric:
 Train:
   dataset:
     name: SimpleDataSet
-    data_dir: /home/admin-root/haotian/康达瑞贝斯机器狗/ocr_dataset/001读表图片
+    data_dir: /home/admin-root/haotian/康达瑞贝斯机器狗/ocr_dataset
     label_file_list:
       - /home/admin-root/haotian/康达瑞贝斯机器狗/ocr_dataset/001读表图片/Label.txt
     ratio_list: [1.0]
@@ -133,13 +133,13 @@ Train:
   loader:
     shuffle: true
     drop_last: false
-    batch_size_per_card: 8
+    batch_size_per_card: 4
-    num_workers: 8
+    num_workers: 4
 
 Eval:
   dataset:
     name: SimpleDataSet
-    data_dir: /home/admin-root/haotian/康达瑞贝斯机器狗/ocr_dataset/001读表图片
+    data_dir: /home/admin-root/haotian/康达瑞贝斯机器狗/ocr_dataset
     label_file_list:
       - /home/admin-root/haotian/康达瑞贝斯机器狗/ocr_dataset/001读表图片/Label.txt
     transforms:

PaddleOCR-3.1.0/configs/rec/PP-OCRv5/PP-OCRv5_server_rec.yml

@@ -5,7 +5,7 @@ Global:
   epoch_num: 75
   log_smooth_window: 20
   print_batch_step: 10
-  save_model_dir: ./output/PP-OCRv5_server_rec
+  save_model_dir: ./output/PP-OCRv5_server_rec_20250815
   save_epoch_step: 1
   eval_batch_step: [0, 2000]
   cal_metric_during_train: true
@@ -78,7 +78,7 @@ Train:
   dataset:
     name: MultiScaleDataSet
     ds_width: false
-    data_dir: ./train_data/
+    data_dir: /home/admin-root/haotian/康达瑞贝斯机器狗/ocr_dataset/001读表图片/crop_img
     ext_op_transform_idx: 1
     label_file_list:
     - ./train_data/train_list.txt

output/3aee64cc1f90d93a5a45979f7b17cb4b_frame_001460_res.json

@@ -0,0 +1,84 @@
+{
+    "input_path": "/home/admin-root/haotian/康达瑞贝斯机器狗/data_image/001读表图片/3aee64cc1f90d93a5a45979f7b17cb4b_frame_001460.jpg",
+    "page_index": null,
+    "model_settings": {
+        "use_doc_preprocessor": true,
+        "use_textline_orientation": false
+    },
+    "doc_preprocessor_res": {
+        "input_path": null,
+        "page_index": null,
+        "model_settings": {
+            "use_doc_orientation_classify": false,
+            "use_doc_unwarping": false
+        },
+        "angle": -1
+    },
+    "dt_polys": [
+        [
+            [
+                813,
+                204
+            ],
+            [
+                868,
+                204
+            ],
+            [
+                868,
+                240
+            ],
+            [
+                813,
+                240
+            ]
+        ]
+    ],
+    "text_det_params": {
+        "limit_side_len": 64,
+        "limit_type": "min",
+        "thresh": 0.3,
+        "max_side_limit": 4000,
+        "box_thresh": 0.6,
+        "unclip_ratio": 1.5
+    },
+    "text_type": "general",
+    "textline_orientation_angles": [
+        -1
+    ],
+    "text_rec_score_thresh": 0.0,
+    "rec_texts": [
+        "1.1"
+    ],
+    "rec_scores": [
+        0.862824022769928
+    ],
+    "rec_polys": [
+        [
+            [
+                813,
+                204
+            ],
+            [
+                868,
+                204
+            ],
+            [
+                868,
+                240
+            ],
+            [
+                813,
+                240
+            ]
+        ]
+    ],
+    "rec_boxes": [
+        [
+            813,
+            204,
+            868,
+            240
+        ]
+    ]
+}