Kornia: differentiable Computer Vision with Pytorch
- Speaker
- Edgar Riba
- Event
- PyDay BCN (2020)
- Date
- Type
- Talk
- Language
- English
- Topic
- Data Science, ML, DL
- Length
- 15 minutes
About this talk
The talk is an introduction to Kornia which is a library that implements classical Computer Vision algorithms in PyTorch that got a lot of attention within the open source community during the last year. The core of the library is written from scratch using pure PyTorch operators and uses the auto-reverse differentiable engine to allow to the user to plug any function within neural networks with a transparent API for high performance devices such as GPUs and TPUs. During the talk I’ll go through the different principle designs and the functionalities found in each of the modules, practical examples, basic usage of the library and more. Information about the project in the website: www.kornia.org