Depth Based Semantic Scene Completion with Position Importance Aware Loss

By Yu Liu*, Jie Li*, Xia Yuan, Chunxia Zhao, Roland Siegwart, Ian Reid and Cesar Cadena (* indicates equal contribution)

ICRA2020 In Conjunction of RAL

Video Demo:

https://youtu.be/j-LAMcMh0yg

Requirements:

python 2.7

pytorch 0.4.1

CUDA 8

Testing

python ./test.py
--data_test=/path/to/NYUCADtest
--batch_size=1
--workers=4
--resume='PALNet_weights.pth.tar'

Weights

Model trained on NYUCAD

Datasets

The original dataset is from SSCNet

Here is the NYUCAD data reproduced from SSCNet for a quick demo.

Adelaide AI Group

more work from Adelaide can be found in: https://github.com/Adelaide-AI-Group/Adelaide-AI-Group.github.io