Update app.py

app.py

@@ -43,7 +43,9 @@ def process_video(video_path, target, progress=gr.Progress()):
         if not ret:
             break
 
-        pil_img = Image.fromarray(cv2.cvtColor(img, cv2.COLOR_BGR2RGB))
+        # Resize the frame to 640x480
+        img_resized = cv2.resize(img, (640, 480))
+        pil_img = Image.fromarray(cv2.cvtColor(img_resized, cv2.COLOR_BGR2RGB))
         
         # Process single image
         inputs = processor(text=[target], images=pil_img, return_tensors="pt", padding=True).to(device)
@@ -56,7 +58,7 @@ def process_video(video_path, target, progress=gr.Progress()):
         max_score = 0
 
         try:
-            font = ImageFont.truetype("arial.ttf", 40)
+            font = ImageFont.truetype("arial.ttf", 20)  # Reduced font size for smaller image
         except IOError:
             font = ImageFont.load_default()
 
@@ -70,7 +72,7 @@ def process_video(video_path, target, progress=gr.Progress()):
                 annotation = f"{object_label}: {confidence}"
 
                 draw.rectangle(box, outline="red", width=2)
-                text_position = (box[0], box[1] - 30)
+                text_position = (box[0], box[1] - 20)
                 draw.text(text_position, annotation, fill="white", font=font)
 
                 max_score = max(max_score, confidence)
@@ -80,7 +82,7 @@ def process_video(video_path, target, progress=gr.Progress()):
 
     cap.release()
     return processed_frames, frame_scores, None
-
+    
 def create_heatmap(frame_scores, current_frame):
     plt.figure(figsize=(12, 3))
     plt.imshow([frame_scores], cmap='hot_r', aspect='auto')  # 'hot_r' for reversed hot colormap