Update configs/pretrain/yolo_world_s_pretrain_FG_v3.py

configs/pretrain/yolo_world_s_pretrain_FG_v3.py

@@ -1,7 +1,7 @@
 # deploy:
 #python deploy/deploy.py /data/taofuyu/models/yolo_world/detection_onnxruntime_static.py /data/taofuyu/models/yolo_world/yolo_world_s_pretrain_FG_v2.py /data/taofuyu/snapshot/yolo_world/fg_pretrain_v2/epoch_1.pth /data/taofuyu/tao_dataset/FG/训练FG_无车顶车窗/192_168_1_123_2_2024-01-15_09-53-05_2024-01-15_09-54-20_0.jpg --test-img /data/taofuyu/tao_dataset/FG/test_wrong/现场问题/第171次车位引导【车头图】识别结果推送_车位2_1_闽A4YY27_None_picture_2023_11_29_18_6_55.jpg --work-dir /data/taofuyu/log/yolo_world/fg_pretrain_v2/
-_base_ = (
-    '/data/taofuyu/repos/YOLO-World/third_party/mmyolo/configs/yolov8/yolov8_s_syncbn_fast_8xb16-500e_coco.py')
+_base_ = ('../../third_party/mmyolo/configs/yolov8/'
+          'yolov8_s_syncbn_fast_8xb16-500e_coco.py')
 custom_imports = dict(
     imports=['yolo_world'],
     allow_failed_imports=False)
@@ -78,56 +78,45 @@ train_pipeline = [
 train_pipeline_stage2 = [*_base_.train_pipeline_stage2[:-1], *text_transform]
 
 
-# part_obj365v1_train_dataset = dict(
-#     type='MultiModalDataset',
-#     dataset=dict(
-#         type='YOLOv5Objects365V1Dataset',
-#         data_root='/data/taofuyu/tao_dataset/yolo_world/objects365v1/',
-#         ann_file='annotations/objects365_train.json',
-#         data_prefix=dict(img='train/'),
-#         filter_cfg=dict(filter_empty_gt=False, min_size=32)),
-#     class_text_path='data/texts/obj365v1_class_texts.json',
-#     pipeline=train_pipeline)
+obj365v1_train_dataset = dict(
+    type='MultiModalDataset',
+    dataset=dict(
+        type='YOLOv5Objects365V1Dataset',
+        data_root='data/objects365v1/',
+        ann_file='annotations/objects365_train.json',
+        data_prefix=dict(img='train/'),
+        filter_cfg=dict(filter_empty_gt=False, min_size=32)),
+    class_text_path='data/texts/obj365v1_class_texts.json',
+    pipeline=train_pipeline)
 
 mg_train_dataset = dict(type='YOLOv5MixedGroundingDataset',
-                        data_root='/data/taofuyu/tao_dataset/yolo_world/mixed_grounding/',
+                        data_root='data/mixed_grounding/',
                         ann_file='annotations/final_mixed_train_no_coco.json',
-                        data_prefix=dict(img='images/'),
+                        data_prefix=dict(img='gqa/images/'),
                         filter_cfg=dict(filter_empty_gt=False, min_size=32),
                         pipeline=train_pipeline)
 
 flickr_train_dataset = dict(
     type='YOLOv5MixedGroundingDataset',
-    data_root='/data/taofuyu/tao_dataset/yolo_world/flickr/',
+    data_root='data/flickr/',
     ann_file='annotations/final_flickr_separateGT_train.json',
-    data_prefix=dict(img='flickr30k_images/'),
+    data_prefix=dict(img='full_images/'),
     filter_cfg=dict(filter_empty_gt=True, min_size=32),
     pipeline=train_pipeline)
 
-fg_train_dataset = dict(
-    type='MultiModalDataset',
-    dataset=dict(
-        type='YOLOv5FGDataset',
-        data_root='',
-        ann_file='/data/taofuyu/tao_dataset/井盖检测/jinggai_few_shot_3.json',
-        data_prefix=dict(img=''),
-        filter_cfg=dict(filter_empty_gt=False, min_size=32)),
-    class_text_path='/data/taofuyu/repos/YOLO-World/data/texts/fewshot_class_texts.json',
-    pipeline=train_pipeline)
-
 train_dataloader = dict(batch_size=train_batch_size_per_gpu,
                         collate_fn=dict(type='yolow_collate'),
                         dataset=dict(_delete_=True,
                                      type='ConcatDataset',
                                      datasets=[
-                                         fg_train_dataset,
+                                         obj365v1_train_dataset,
                                          flickr_train_dataset, mg_train_dataset
                                      ],
                                      ignore_keys=['classes', 'palette']))
 
 test_pipeline = [
     *_base_.test_pipeline[:-1],
-    dict(type='LoadText', text_path='/data/taofuyu/repos/YOLO-World/data/texts/fewshot_class_texts.json'),
+    dict(type='LoadText'),
     dict(type='mmdet.PackDetInputs',
          meta_keys=('img_id', 'img_path', 'ori_shape', 'img_shape',
                     'scale_factor', 'pad_param', 'texts'))
@@ -135,23 +124,20 @@ test_pipeline = [
 coco_val_dataset = dict(
     _delete_=True,
     type='MultiModalDataset',
-    dataset=dict(
-        type='YOLOv5FGDataset',
-        data_root='',
-        ann_file='/data/taofuyu/tao_dataset/井盖检测/jinggai_few_shot_3.json',
-        data_prefix=dict(img=''),
-        filter_cfg=dict(filter_empty_gt=False, min_size=32)),
-    class_text_path='/data/taofuyu/repos/YOLO-World/data/texts/fewshot_class_texts.json',
+    dataset=dict(type='YOLOv5LVISV1Dataset',
+                 data_root='data/coco/',
+                 test_mode=True,
+                 ann_file='lvis/lvis_v1_val.json',
+                 data_prefix=dict(img=''),
+                 batch_shapes_cfg=None),
+    class_text_path='data/texts/lvis_v1_class_texts.json',
     pipeline=test_pipeline)
 val_dataloader = dict(dataset=coco_val_dataset)
 test_dataloader = val_dataloader
 
-val_evaluator = dict(
-    _delete_=True,
-    type='mmdet.CocoMetric',
-    proposal_nums=(100, 1, 10),
-    ann_file='/data/taofuyu/tao_dataset/井盖检测/jinggai_few_shot_3.json',
-    metric='bbox')
+val_evaluator = dict(type='mmdet.LVISMetric',
+                     ann_file='data/coco/lvis/lvis_v1_val.json',
+                     metric='bbox')
 test_evaluator = val_evaluator
 
 # training settings
@@ -170,7 +156,7 @@ custom_hooks = [
          switch_pipeline=train_pipeline_stage2)
 ]
 train_cfg = dict(max_epochs=max_epochs,
-                 val_interval=100,
+                 val_interval=10,
                  dynamic_intervals=[((max_epochs - close_mosaic_epochs),
                                      _base_.val_interval_stage2)])
 optim_wrapper = dict(optimizer=dict(