output top 3 results

app.py

@@ -45,8 +45,12 @@ def search_in_video(video, query):
 
     ordered_similarity_scores = similarities[0][arg_sorted_similarities]
 
-    best_video = video_segments[arg_sorted_similarities[0, -1]]
-    torchvision.io.write_video('best.mp4', best_video, video_fps)
+    top1 = video_segments[arg_sorted_similarities[0, -1]]
+    top2 = video_segments[arg_sorted_similarities[0, -2]]
+    top3 = video_segments[arg_sorted_similarities[0, -3]]
+    torchvision.io.write_video('top1.mp4', top1, video_fps)
+    torchvision.io.write_video('top2.mp4', top2, video_fps)
+    torchvision.io.write_video('top3.mp4', top3, video_fps)
 
     total_frames = video.shape[0]
 
@@ -70,11 +74,16 @@ def search_in_video(video, query):
         zip(ordered_start_ends[0].tolist(), ordered_similarity_scores[0].tolist())
     )
 
-    return 'best.mp4', labels_to_scores
+    return 'top1.mp4', 'top2.mp4', 'top3.mp4', labels_to_scores
 
 app = gr.Interface(
     fn=search_in_video,
     inputs=['video', 'text'],
-    outputs=['video', gr.outputs.Label(num_top_classes=3, type='auto')],
+    outputs=[
+        gr.Video(format='mp4', label='Top1'),
+        gr.Video(format='mp4', label='Top2'),
+        gr.Video(format='mp4', label='Top3'),
+        gr.outputs.Label(num_top_classes=5, type='auto', label='Scores'),
+    ],
 )
 app.launch()