Update app.py

app.py

@@ -64,15 +64,24 @@ def predict_with_beta_vae(sequence, num_frames):
 def generate_predictions(frame_list, model_choice, mode_choice, num_frames):
     """
     Args:
-        frame_list: uploaded images (list of PIL)
+        frame_list: uploaded images (list of file paths as strings)
         model_choice: 'ConvLSTM' or 'β-VAE'
         mode_choice: 'Single Frame' or 'Multiple Frames'
         num_frames: number of consecutive frames to generate
     """
-    if len(frame_list) < SEQUENCE_LENGTH:
+    if frame_list is None or len(frame_list) < SEQUENCE_LENGTH:
         raise gr.Error(f"Please upload at least {SEQUENCE_LENGTH} sequential frames.")
 
-    frames = frame_list[:SEQUENCE_LENGTH]
+    # Convert file paths to PIL Images
+    frames = []
+    for file_path in frame_list[:SEQUENCE_LENGTH]:
+        # Handle both string paths and file objects
+        if isinstance(file_path, str):
+            img = Image.open(file_path)
+        else:
+            img = Image.open(file_path.name)
+        frames.append(img)
+    
     processed = [preprocess_frame(f) for f in frames]
     sequence = torch.cat(processed, dim=0).unsqueeze(0)  # (1, T, 1, H, W)
 
@@ -92,7 +101,6 @@ def generate_predictions(frame_list, model_choice, mode_choice, num_frames):
 description = """
 # 🕹️ Pong Frame Prediction  
 Upload **10 sequential Pong frames** and select a model + prediction mode.
-
 - **ConvLSTM** → Learns temporal dynamics directly in pixel space  
 - **β-VAE** → Predicts next frames via latent-space reconstruction  
 """
@@ -105,9 +113,7 @@ demo = gr.Interface(
         gr.Radio(["Single Frame", "Multiple Frames"], label="Prediction Mode", value="Single Frame"),
         gr.Slider(1, 20, value=5, step=1, label="Number of Consecutive Frames (if Multiple Mode)"),
     ],
-
-    outputs = gr.Gallery(label="Predicted Frames", elem_id="predicted-frames",show_label=True, columns=2), # columns replaces grid
-
+    outputs=gr.Gallery(label="Predicted Frames", elem_id="predicted-frames", show_label=True, columns=2),
     title="Pong Frame Predictor (ConvLSTM / β-VAE)",
     description=description,
 )
@@ -116,4 +122,4 @@ demo = gr.Interface(
 # Launch App
 # ===============================================================
 if __name__ == "__main__":
-    demo.launch()
+    demo.launch()