- Clone the repo
- Open the command line and navigate to the repo
- Type pip install -r requirements.txt and enter
- Download supported pre-trained models RetinaNet / YOLOv3 / TinyYOLOv3 at Models for Image Recognition and Object Detection and put them to the repo. Now you can detect and recognize 80 different kind of common everyday objects in any video
- To train you own model to recognize everything you want, please refer to Custom Training with ImageAI
- Author of ImageAI: https://github.com/OlafenwaMoses/ImageAI
- If you run this on CPU, you can only test with a very short video. Trust me, it took me 4 hrs to proceed a 10s video on my potato one. Otherwise, just buy GPU 😁
- Video source: France vs Croatia WC 2018 Final