uploaded videos from users

app.py

@@ -84,7 +84,34 @@ def images_to_video(image_list, output_path, fps=10):
 
     writer.close()
 
+def video2frames(video_path):
+    # Open the video file
+    video = cv2.VideoCapture(video_path)
 
+    img_path = video_path[:-4]
+    # Initialize frame counter
+    frame_count = 0
+    os.makedirs(img_path, exist_ok=True)
+
+    while True:
+        # Read a frame from the video
+        ret, frame = video.read()
+
+        # If the frame was not successfully read, then we have reached the end of the video
+        if not ret:
+            break
+
+        # Write the frame to a JPG file
+        frame_file = f"{img_path}/{frame_count:05}.jpg"
+        cv2.imwrite(frame_file, frame)
+
+        # Increment the frame counter
+        frame_count += 1
+
+    # Release the video file
+    video.release()
+
+    return img_path
 
 @spaces.GPU(duration=120)
 def DiffBIR_restore(input_video, prompt, sr_ratio, n_frames, n_steps, guidance_scale, seed, n_prompt, task):
@@ -93,7 +120,8 @@ def DiffBIR_restore(input_video, prompt, sr_ratio, n_frames, n_steps, guidance_s
     if video_name in video_to_image:
         frames_path = video_to_image[video_name][0]
     else:
-        return None
+        frames_path = video2frames(input_video)
+
     print(f"[INFO] input_video: {input_video}")
     print(f"[INFO] Frames path: {frames_path}")
     args = argparse.Namespace()
@@ -195,7 +223,7 @@ with gr.Blocks(css="style.css") as demo:
             n_steps = gr.Slider(label='Steps',
                                     minimum=1,
                                     maximum=100,
-                                    value=10,
+                                    value=5,
                                     step=1)
             guidance_scale = gr.Slider(label='Guidance Scale',
                                         minimum=0.1,
@@ -262,7 +290,7 @@ with gr.Blocks(css="style.css") as demo:
             n_steps = gr.Slider(label='Steps',
                                     minimum=1,
                                     maximum=100,
-                                    value=10,
+                                    value=5,
                                     step=1)
             guidance_scale = gr.Slider(label='Guidance Scale',
                                         minimum=0.1,

utils/batch_inference.py

@@ -148,7 +148,6 @@ class InferenceLoop:
     
     def batch_lq_loader(self) -> Generator[np.ndarray, None, None]:
         img_exts = [".png", ".jpg", ".jpeg"]
-        print(f"[INFO] input: {self.args.input}")
         if os.path.isdir(self.args.input):
             file_names = sorted([
                 file_name for file_name in os.listdir(self.args.input) if os.path.splitext(file_name)[-1] in img_exts