RoadNet-RT

Lightweight and Real-Time Road Segmentation Network

Requirements

• python==3.6
• keras==2.3.1
• tensorflow-gpu==1.14.0
• opencv=3.4.2
• matplotlib==3.1.1
• albumentations==0.4.5
• h5py==2.10.0

How to use

Step 0

Donwload KITTI Dataset for Road Segmentation and put it in the path "../". Or you can modify the dataset path in roadnet_train.py and roadnet_test.py.

Step 1

Install the enviroment according to the requirement above.

Step 2

Install Segmentation Models

$ pip install -U segmentation-models

Step 3

If inference, run the command bellow

$ python roadnet_test.py

Result on KITTI leaderboard

The inference speed on GTX1080 is about 9ms/frame, with accuracy 92.55%. Link to KITTI.

Citation

@ARTICLE{roadnet-rt2021,
  author={Bai, Lin and Lyu, Yecheng and Huang, Xinming},
  journal={IEEE Transactions on Circuits and Systems I: Regular Papers}, 
  title={RoadNet-RT: High Throughput CNN Architecture and SoC Design for Real-Time Road Segmentation}, 
  year={2021},
  volume={68},
  number={2},
  pages={704-714},
  doi={10.1109/TCSI.2020.3038139}}

Acknowledgement

This repo is heavily relied on Segmentation Models. We thank Pavel Yakubovskiy for his great work.