jeou / MXNetSeg

This project aims to provide modular implementation and easy training pipeline for SOTA semantic segmentation models.

MXNetSeg

This project provides modular implementation for state-of-the-art semantic segmentation models based on MXNet framework and GluonCV toolkit.

Environment

We adopt python 3.6.2 and CUDA 10.1 in this project.

1. prerequisites

   pip install -r requirements.txt

2. Detail API for Pascal Context dataset

3. 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

Usage

Training

1. Configure hyper-parameters in ./mxnetseg/models/$MODEL_NAME$/config.py

2. Run the train.py script

   python train.py --model fcn --ctx 0 1 2 3 --val

Inference

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

Citations

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}
}