"# 2022-U-Net" "# 2022-U-Net" "# 2022-U-Net"

This repository apply the U-Net on the KITTI semantic segementation dataset. The GitHub action works well, but it does not store and write the folder to generate semantic segmentation images. If you would like to do pull request, could you please download the train image, train semantic, test images and hdf5 file to see the entire thing and modify the code. The datasets are small.

Reference:

http://www.cvlibs.net/datasets/kitti/eval_semantics.php

https://arxiv.org/abs/1505.04597

BCDU-net: Bi-directional ConvLSTM U-net (BCDU-net) is a deep learning-based network and an extension of the U-net. It works by taking full advantages of U-Net, bi-directional ConvLSTM (BConvLSTM) and the mechanism of dense convolutions. This network was used for the segmentation of medical images and I was curious how it will perform on this dataset. The results weren't very accurate but maybe if someone is interested they can look further into this neural network and even use it for other types od dataset.