This project provides modular implementation for state-of-the-art semantic segmentation models based on MXNet framework and GluonCV toolkit.
We adopt python 3.6.2 and CUDA 10.1 in this project.
-
prerequisites
pip install -r requirements.txt
-
Detail API for Pascal Context dataset
-
We also employ fitlog to generate training logs in addition to the *logging package. Run the following command to initialize this project to a fitlog one
fitlog init
-
Configure hyper-parameters in ./mxnetseg/models/$MODEL_NAME$/config.py
-
Run the
train.py
scriptpython train.py --model fcn --ctx 0 1 2 3 --val
Simply run the eval.py
with arguments need to be specified
python eval.py --model fcn --backbone resnet50 --resume ./weights/fcnfcn_resnet18_Cityscapes_05_19_00_31_06_best.params --ctx 0 --data cityscapes --ms
Please kindly cite our paper if you feel our codes helps in your research.
@article{tang2020attention,
title={Attention-guided Chained Context Aggregation for Semantic Segmentation},
author={Tang, Quan and Liu, Fagui and Jiang, Jun and Zhang, Yu},
journal={arXiv preprint arXiv:2002.12041},
year={2020}
}