Skip to content
master
Switch branches/tags
Go to file
Code

Latest commit

s
a6d2d4c

Git stats

Files

Permalink
Failed to load latest commit information.
Type
Name
Latest commit message
Commit time
s
Jul 31, 2019
s
Jul 31, 2019
s
Jul 31, 2019
s
Jul 31, 2019
s
Jul 31, 2019
s
Jul 31, 2019

README.md

This repository includes the evaluation code for the paper "Regularizing Proxies with Multi-Adversarial Training for Unsupervised Domain-Adaptive Semantic Segmentation (Arxiv Link)". The whole package will be made public soom.

Dependencies:

mxnet

gluoncv

numpy

tqdm

easydict

yaml

pillow

Dataset:

To evaluate the performance on Cityscapes dataset, please first put the dataset into the correct path. Please edit the variable "data_root" in "cfg/resnet101_gta2cs.yaml" and "cfg/resnet101_syn2cs.yaml", which points to the data root. Then name the Cityscapes folder "cityscapes" and put it in the data root.

Evaluation:

We provide two resnet101 models for GTAV->CS and SYNTHIA->CS respectively. first download the models HERE and HERE:

To evaluate them, simply run:

python eval.py --cfg cfg/resnet101_gta2cs.yaml --resume resnet101_gta2cs.params

python eval.py --cfg cfg/resnet101_syn2cs.yaml --resume resnet101_syn2cs.params

About

No description, website, or topics provided.

Resources

Releases

No releases published

Packages

No packages published

Languages