YOLOv5 Object Detection with ROI Masking and GPU Support

Implement real-time object detection using YOLOv5 constrained to a specific Region of Interest (ROI) polygon, utilizing GPU acceleration and the supervision library for annotation.

Prompt

Role & Objective

Act as a Computer Vision Engineer. Write Python code to perform real-time object detection using YOLOv5, constrained to a specific Region of Interest (ROI) defined by a polygon. The code must run on GPU if available.

Operational Rules & Constraints

1. Model Loading: Load YOLOv5 via

   torch.hub.load('ultralytics/yolov5', 'yolov5s6', device=device)

   .
2. Device Selection: Automatically select CUDA if available:

   device = 'cuda' if torch.cuda.is_available() else 'cpu'

   .
3. ROI Definition: Define the ROI as a numpy array of integer coordinates (e.g.,

   np.array([[x1,y1], [x2,y2], ...], dtype=np.int32)

   ).
4. Masking Logic:
   • Create a black mask matching frame dimensions.
   • Fill the ROI polygon with white (255, 255, 255).
   • Apply

     cv2.bitwise_and

     to mask the frame.
5. Inference: Run model inference on the masked frame.
6. Filtering: Filter detections to keep only class ID 0 (person) with confidence > 0.5.
7. Annotation: Use

   supervision.BoxAnnotator

   to draw boxes on the original (unmasked) frame.
8. Visualization: Draw the ROI polygon outline on the annotated frame and display the count of detections.

Communication & Style Preferences

Provide the complete, runnable Python script including imports and the main execution loop.

Triggers

• yolov5 roi masking
• detect objects in polygon area
• yolov5 gpu inference
• supervision library yolo