xmba15/enet_pytorch

fast semantic segmentation with Enet

Jupyter Notebook

Pytorch Implementation of ENet: A Deep Neural Network Architecture for Real-Time Semantic Segmentation

TODO

Implement Enet network
Train Enet on Camvid Dataset
Train Enet on Cityscapes Dataset
Test with Lovasz Softmax Loss

Camvid Dataset

This dataset consists of street images whose pixels are labeled into 11+1 classes (one "Void" class for the ego-car)
Run

bash ./scripts/download_camvid.sh

to fetch and unzip the dataset into data directory

Colors for Each Label Class

How To

Train on Camvid:

python3 camvid_train.py

Test on test set of Camvid

python3 ./scripts/test_camvid_testset.py

You should see the following result:

References