Real-time object detection using a custom-trained YOLOv3 model based on the COCO dataset, integrated with OpenCV for video and image inference.
This project implements real-time object detection using a YOLOv3 model trained on the COCO dataset. It utilizes OpenCV's DNN module for fast inference and supports image, video, and webcam detection.
ai_model/
├── yolov3.cfg # YOLOv3 configuration file
├── yolov3.weights # Trained weights (COCO)
├── coco.names # COCO class labels (80 classes)
├── detect.py # Python script for running inference
##🧠 Model Info
Model Type: YOLOv3
Dataset: COCO (Common Objects in Context)
Classes: 80 (person, bicycle, car, dog, etc.)
Framework: OpenCV with DNN backend
Input Size: 416×416
##🔧 Setup
Install dependencies:
bash
Copy
Edit
pip install opencv-python numpy
Download YOLOv3 weights if not present:
https://pjreddie.com/media/files/yolov3.weights
Run the detection script:
bash
Copy
Edit
python detect.py --image input.jpg
Or for webcam:
bash
Copy
Edit
python detect.py --webcam