ENet - Real Time Semantic Segmentation

bash grab.sh # fetch runtime deps

A Neural Net Architecture for real time Semantic Segmentation.
In this repository we have reproduced the ENet Paper - Which can be used on mobile devices for real time semantic segmentattion. The link to the paper can be found here: ENet

How to use?

This repository comes in with a handy notebook which you can use with Colab.
You can find a link to the notebook here: ENet - Real Time Semantic Segmentation
Open it in colab: Open in Colab

Clone the repository and cd into it

git clone https://github.com/iArunava/ENet-Real-Time-Semantic-Segmentation.git
cd ENet-Real-Time-Semantic-Segmentation/

Use this command to train the model

python3 init.py --mode train -iptr path/to/train/input/set/ -lptr /path/to/label/set/

Use this command to test the model

python3 init.py --mode test -m /path/to/the/pretrained/model.pth -i /path/to/image/to/infer.png

Use --help to get more commands

python3 init.py --help

Some results

enet infer 1 enet infer 4 enet infer 6 enet infer 5 enet infer 2

References

A. Paszke, A. Chaurasia, S. Kim, and E. Culurciello. Enet: A deep neural network architecture for real-time semantic segmentation. arXiv preprint arXiv:1606.02147, 2016.

Citations

@inproceedings{ BrostowSFC:ECCV08,
  author    = {Gabriel J. Brostow and Jamie Shotton and Julien Fauqueur and Roberto Cipolla},
  title     = {Segmentation and Recognition Using Structure from Motion Point Clouds},
  booktitle = {ECCV (1)},
  year      = {2008},
  pages     = {44-57}
}

@article{ BrostowFC:PRL2008,
    author = "Gabriel J. Brostow and Julien Fauqueur and Roberto Cipolla",
    title = "Semantic Object Classes in Video: A High-Definition Ground Truth Database",
    journal = "Pattern Recognition Letters",
    volume = "xx",
    number = "x",   
    pages = "xx-xx",
    year = "2008"
}

License

The code in this repository is distributed under the BSD v3 Licemse.
Feel free to fork and enjoy :)

iArunava/ENet-Real-Time-Semantic-Segmentation

計測したスター増加

概要

README.md