Skip to content

Unofficial implementation of CARAFE: Content-Aware ReAssembly of FEatures. Pure pytorch imp

Notifications You must be signed in to change notification settings

XiaLiPKU/CARAFE

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

5 Commits
 
 
 
 

Repository files navigation

CARAFE

An unofficial implementation of CARAFE: Content-Aware ReAssembly of FEatures

Usage

Download the raw file of carafe.py into your project, and then import it by: from carafe import CARAFE

Some results

By now, I've only experimented on the Sementic Segmentation task. The results are reported on the Cityscapes dataset. The backbone is ResNet-101 with output stride 32 (no dilation is adopted). For more details, please refer Table 6 in the original paper. PPM and FUSE are not adopted here. I only compare upon FPN here.

Methods mIoU
Bilinear 74.52
CARAFE(k=3) 78.16
CARAFE(k=5) 78.82

About

Unofficial implementation of CARAFE: Content-Aware ReAssembly of FEatures. Pure pytorch imp

Resources

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published

Languages