icnet

This repository contains the TensorFlow implementation for the following paper

ICNet for RT Semantic Segmentation (ECCV_2018)

if you use this code for your research, please consider citing:

@inProceedings{
      title={ICNet for Real-Time Semantic Segmentation on High-Resolution Images},
      author={Hengshuang Zhao1, Xiaojuan Qi1, Xiaoyong Shen2, Jianping Shi3, Jiaya Jia1,2},
      booktitle={ECCV},
      year={2018}
   }

Project Page

Dependencies

Requirements:

• python 3.6 with Numpy and opencv-python
• tensorflow (version 1.0+)
• etc

My code has been tested with python 3.6, tensorflow 1.13.0, CUDA 11.3 on Window10

Runing the demo

Installation

Dataset

I used the cityscapes-dataset When using the provided data make sure to respect the cityscapes-dataset license.

• https://www.cityscapes-dataset.com/
• The cityscapes dataset for semantic urban understanding, arXiv: 1604.01685

Below is the complete set of training data. Download it into the data/ forder

• https://drive.google.com/drive/folders/1Wjr7-V0QZMTd-yDf_IuoJ0sYoRTDfyBW?usp=sharing

• 구글드라이브

• data directory:

  
  cityscapes
    |-------leftImg8bit---train
    |-------gtFine -------train
  

Training

 python train.py 

You can change the training data, learning rate and other parameters by editing train.py The total number of training epochs is 100 ; learning rate is initialized as 1e-3 and training epoch of 100 with linear decay after 50 epoches

Evaluation

Statement

Contact

Kim Min Geyung Email: muzzcats@naver.com

License