shounaknaik / semantic-point-painting Goto Github PK

To run this project follow these steps:

1. Run the 2 download shell script files. This will download the particular sequence of KITTI-360.
2. Convert the binary files to pcd with the bin2pcd.py in Code folder.
3. Make an map using ICP by running the icp.py file in Code/pcd_folder folder.
4. Semantically segment images by running semantic_segmentation.py in DeepLabV3Plus-Pytorch folder.
5. Again generate the map using icp.py

Alternatively , run the run.bash script which will everything for you.

NOTE: Download the pretrained weights for the network from https://www.dropbox.com/sh/w3z9z8lqpi8b2w7/AAABmvcqWdVtTJCFQ75OmK0va/best_deeplabv3_resnet101_voc_os16.pth?dl=0 and store them in the Code folder. Keep the first 50 frames from the dowloaded sequence for our project.