Update app.py

app.py

@@ -157,9 +157,11 @@ def make_video(video_path, outdir='./vis_video_depth', encoder='vits'):
             # out.write(combined_frame)
             # frame_path = os.path.join(temp_frame_dir, f"frame_{count:05d}.png")
             # cv2.imwrite(frame_path, combined_frame)
-            
+
             cv2.imwrite(f"f{count}.jpg", raw_frame)
             orig_frames.append(f"f{count}.jpg")
+            global masks
+            masks.append(f"f{count}.jpg")
             cv2.imwrite(f"f{count}_dmap.jpg", depth_color)
             depth_frames.append(f"f{count}_dmap.jpg")
             count += 1
@@ -171,12 +173,10 @@ def make_video(video_path, outdir='./vis_video_depth', encoder='vits'):
         cv2.destroyAllWindows()
 
         global frame_selected
-        global masks
         global depths
         global frames
         frames = orig_frames
         depths = depth_frames
-        masks = orig_frames
         return final_vid, final_zip, frames, masks[frame_selected] #output_path
 
 def depth_edges_mask(depth):
@@ -379,7 +379,7 @@ with gr.Blocks(css=css) as demo:
             input_url = gr.Textbox(value="./examples/streetview.mp4", label="URL")
             input_video = gr.Video(label="Input Video", format="mp4")
             input_url.change(fn=loadurl, inputs=[input_url], outputs=[input_video])
-            output_frame = gr.Gallery(label="Frames", preview=True, columns=8192)
+            output_frame = gr.Gallery(label="Frames", preview=True, columns=8192, type="filepath")
             output_switch = gr.Checkbox(label="Show depths")
             output_switch.input(fn=switch_rows, inputs=[output_switch], outputs=[output_frame])
             output_mask = gr.ImageEditor(interactive=True, transforms=(None,), eraser=gr.Eraser(), brush=gr.Brush(colors=['black', 'darkgray', 'gray', 'lightgray', 'white']), layers=True)