Deep Learning and Applications summer course (2021, FGV). Professor: Cristian Jose Lopez Del Alamo (Universidad la Salle - Peru).
In this couse, we studied various neural networks architectures, focusing on the implementation, such as CNNs, RNNs, Autoencoders, Variational Autoencoders and GANs. In this repository, you will find implementations of
- CNN for MNIST (notebook)
- CNN architecture comparison for COVID-19 detection (notebook, YouTube)
- social distancing detection with YOLOv5 (notebook, YouTube)
I used the model YOLOv5 for the people detection task and I implemented two ways of calculating distances among people:
- The first (naive) way is to calculate the distance among people using the distance among the centers of the detection boxes returned by the model.
- The second way, cooler, is to apply perspective in order to calculate the real distance among people.
More details can be found in the notebook and on YouTube video. Example of perspective transformation:
