Hi everyone!

My team and I have just released a computer vision algorithm that can detect “thumb up” and “thumb down” gestures.

The algorithm is based on tiny-YOLOv3 architecture. Accuracy of thumb up/down gesture recognition is calculated as mean average precision (mAP@0.25) = 85.19%; average IoU = 73.89%.

The neural network was trained on 3000 images.

To try out the algorithm, download it from the GitHub and install it.

We would like to hear your feedback so we can continue to improve our service to you 🙂