Fine tune a RetinaNet to create a custom model.

Wouldn’t it be frustrating if your smartphone needed to have thousands of pictures of you to recognize you and get unlocked? Thanks for the few-shot learning, this is not needed.

This technique has drawn a lot of attention in the research community and many solutions have been developed. To predict something based on a few training examples, the solutions right now use meta-learning or in three words: learning to learn.

RetinaNet is one of the most used few-shot learning convolution neural networks. In this repo, we are going to use TensorFlow and Python to fine tune this architecture and train a custom model.

If you want to learn how the few-shot detectors work, open the Few Shot Learning: RetinaNet.ipynb notebook and follow the steps to create your own object detector and run it in real-time.

This repository is part of an Expert Class in the Analytics Academy - powered by the Growth Analytics Center, AmbevTech and the BudLab at Ab InBev