Implement people detection in a few minutes

Using the ready-to-use cvlib library to use pre-trained models

This is a 3 parts series where I play with cvlib:

1. Implement people detection in a few minutes
2. Implement face detection in a few minutes
3. Censor faces in a video

I’m getting into Machine Learning and as every time I try something new in code, I like to play a little before I can start my more serious projects. This enables me to have fun learning how a technology or language works.

Here, I decided to try Machine Vision with Python. I studied how Neural Networks are supposed to work and how to train a model and deploy it to use inference on a video to detect object. It is complicated so I decided to go slowly and use cvlib where the models are pre-trained for Yolo-v3 so I don’t have to worry about model training. Here, I’m going to use Yolo-v3-tiny to have faster performances.

Cvlib is a library wrapped around opencv maintained by Arun Ponnusamy, a Computer Vision Reasearch Engineer from India.

You can see it documented here and I will be using the example codes he hosts on GitHub as a base.

I’m documenting here my tests so if you are getting into all this like me, you can follow, do the same and maybe learn too. I hope this can help you but keep in mind that I’m new to ML but also new to Python (I come from C# and Javascript…