Home

Awesome

RoIAlign for PyTorch

This is a PyTorch version of RoIAlign. This implementation is based on crop_and_resize and supports both forward and backward on CPU and GPU.

NOTE: Thanks meikuam for updating this repo for PyTorch 1.0. You can find the original version for torch <= 0.4.1 in pytorch_0.4 branch.

Introduction

The crop_and_resize function is ported from tensorflow, and has the same interface with tensorflow version, except the input feature map should be in NCHW order in PyTorch. They also have the same output value (error < 1e-5) for both forward and backward as we expected, see the comparision in test.py.

Note: Document of crop_and_resize can be found here. And RoIAlign is a wrap of crop_and_resize that uses boxes with unnormalized (x1, y1, x2, y2) as input (while crop_and_resize use normalized (y1, x1, y2, x2) as input). See more details about the difference of RoIAlign and crop_and_resize in tensorpack.

Warning: Currently it only works using the default GPU (index 0)

Usage