RF-DETR Video Object Detection Tool

This project uses the RF-DETR model to detect specific objects in a video. It generates two outputs based on the detected objects:

• 🔳 Cropped Video: A video composed only of cropped frames containing the detected objects.
• 🟥 Annotated Video: The original video with bounding boxes drawn around detected objects.

---

🛠️ Requirements

You need the following Python packages installed:

pip install -r requirements.txt

---

python main.py
--video-path "videos/input_video.mp4"
--cropped-output-path "outputs/cropped_output.mp4"
--annotated-output-path "outputs/annotated_output.mp4"
--class-id 1
--threshold 0.5
--fps 25
--padding 20

---

Argument	Description
--video-path	Path to the input video file (required)
--cropped-output-path	Output path for the cropped objects video
--annotated-output-path	Output path for the annotated full-frame video
--class-id	COCO class ID to detect (e.g. person = 1)
--threshold	Confidence threshold for detection (default: 0.5)
--fps	Frames per second for output videos
--padding	Padding in pixels around bounding boxes when cropping

outputs/
├── cropped_output.mp4       # Video with only cropped detected objects
└── annotated_output.mp4     # Original video with bounding boxes drawn