Skip to content

Latest commit

 

History

History
15 lines (10 loc) · 801 Bytes

File metadata and controls

15 lines (10 loc) · 801 Bytes

Vision Research Toolkit by PyTorch

Why should we build a new framework every time starting a new research?

Welcome to VRT by PyTorch!

  • VRT is a high robust PyTorch-based framework for deep learning in computer vision, which aimed at easy-to-use, high efficiency, clean code and research.
  • The dataset in VRT is efficient enough to support all kind of task (seg, det, class, etc..)
  • All the parameters in VRT can be controlled in one config file.
  • VRT support single GPU and multi GPU, but do not support CPU.

Acknowledgement

  • The idea was inspired by the project torchcv.
  • The base framework is heavily influenced by maskrcnn-benchmark, which is a really cool implementation of PyTorch.