[to #42322933]adjust output form

adjust output form for movie scene segmentation demo

 Link: https://code.alibaba-inc.com/Ali-MaaS/MaaS-lib/codereview/10244194

modelscope/models/cv/movie_scene_segmentation/model.py

@@ -162,11 +162,11 @@ class MovieSceneSegmentationModel(TorchModel):
         thres = self.cfg.pipeline.save_threshold
 
         anno_dict = get_pred_boundary(pred_dict, thres)
-        scene_dict, scene_list = pred2scene(self.shot2keyf, anno_dict)
+        scene_dict_lst, scene_list = pred2scene(self.shot2keyf, anno_dict)
         if self.cfg.pipeline.save_split_scene:
             re_dir = scene2video(inputs['input_video_pth'], scene_list, thres)
             print(f'Split scene video saved to {re_dir}')
-        return len(scene_list), scene_dict
+        return len(scene_list), scene_dict_lst
 
     def preprocess(self, inputs):
         logger.info('Begin shot detect......')

modelscope/models/cv/movie_scene_segmentation/utils/save_op.py

@@ -21,16 +21,15 @@ def get_pred_boundary(pred_dict, threshold=0.5):
 def pred2scene(shot2keyf, anno_dict):
     scene_list, pair_list = get_demo_scene_list(shot2keyf, anno_dict)
 
-    scene_dict = {}
+    scene_dict_lst = []
     assert len(scene_list) == len(pair_list)
     for scene_ind, scene_item in enumerate(scene_list):
-        scene_dict.update(
-            {scene_ind: {
-                'shot': pair_list[scene_ind],
-                'frame': scene_item
-            }})
+        scene_dict_lst.append({
+            'shot': pair_list[scene_ind],
+            'frame': scene_item
+        })
 
-    return scene_dict, scene_list
+    return scene_dict_lst, scene_list
 
 
 def scene2video(source_movie_fn, scene_list, thres):

modelscope/outputs.py

@@ -38,7 +38,7 @@ class OutputKeys(object):
     HISTORY = 'history'
     TIMESTAMPS = 'timestamps'
     SPLIT_VIDEO_NUM = 'split_video_num'
-    SPLIT_META_DICT = 'split_meta_dict'
+    SPLIT_META_LIST = 'split_meta_list'
 
 
 TASK_OUTPUTS = {
@@ -293,18 +293,17 @@ TASK_OUTPUTS = {
     # movide scene segmentation result for a single video
     # {
     #        "split_video_num":3,
-    #        "split_meta_dict":
-    #        {
-    #           scene_id:
+    #        "split_meta_list":
+    #        [
     #           {
     #               "shot": [0,1,2],
     #               "frame": [start_frame, end_frame]
     #           }
-    #        }
+    #        ]
     #
     # }
     Tasks.movie_scene_segmentation:
-    [OutputKeys.SPLIT_VIDEO_NUM, OutputKeys.SPLIT_META_DICT],
+    [OutputKeys.SPLIT_VIDEO_NUM, OutputKeys.SPLIT_META_LIST],
 
     # ============ nlp tasks ===================
 

modelscope/pipelines/cv/movie_scene_segmentation_pipeline.py

@@ -60,9 +60,9 @@ class MovieSceneSegmentationPipeline(Pipeline):
 
     def postprocess(self, inputs: Dict[str, Any]) -> Dict[str, Any]:
         data = {'input_video_pth': self.input_video_pth, 'feat': inputs}
-        video_num, meta_dict = self.model.postprocess(data)
+        video_num, meta_lst = self.model.postprocess(data)
         result = {
             OutputKeys.SPLIT_VIDEO_NUM: video_num,
-            OutputKeys.SPLIT_META_DICT: meta_dict
+            OutputKeys.SPLIT_META_LIST: meta_lst
         }
         return result