mirrors
/
SegFormer
mirror of https://github.com/NVlabs/SegFormer.git
1
0
Fork 0
Code Issues Packages Projects Releases Wiki Activity
Official PyTorch implementation of SegFormer
1 Commit
1 Branch
0 Tags
3.2 MiB
Python 99.8%
Shell 0.1%
 
 
Go to file
xieenze 073b02d986 update 2021-06-13 00:32:37 +08:00
configs update 2021-06-13 00:32:37 +08:00
demo update 2021-06-13 00:32:37 +08:00
docker update 2021-06-13 00:32:37 +08:00
docs update 2021-06-13 00:32:37 +08:00
local_configs update 2021-06-13 00:32:37 +08:00
mmseg update 2021-06-13 00:32:37 +08:00
requirements update 2021-06-13 00:32:37 +08:00
resources update 2021-06-13 00:32:37 +08:00
tests update 2021-06-13 00:32:37 +08:00
tools update 2021-06-13 00:32:37 +08:00
.gitignore update 2021-06-13 00:32:37 +08:00
LICENSE update 2021-06-13 00:32:37 +08:00
README.md update 2021-06-13 00:32:37 +08:00
pytest.ini update 2021-06-13 00:32:37 +08:00
requirements.txt update 2021-06-13 00:32:37 +08:00
setup.cfg update 2021-06-13 00:32:37 +08:00
setup.py update 2021-06-13 00:32:37 +08:00

README.md

SegFormer: Simple and Efficient Design for Semantic Segmentation with Transformers

We use MMSegmentation v0.13.0 as the codebase.

How to install

Install according to the guidelines in MMSegmentation v0.13.0.

Data preparation

Prepare ADE20K, Cityscapes according to the guidelines in MMSegmentation v0.13.0.

Evaluation

First, download trained weights from google drive. Here we provide weights of SegFormer-B1 on ADE20K.

For example, to evaluate SegFormer-B1 on ADE20K on a single node with 8 gpus run:

./tools/dist_test.sh local_configs/segformer/B1/segformer.b1.512x512.ade.160k.py /path/to/checkpoint_file 8