CodeSLAM

Unofficial implement of "CodeSLAM — Learning a Compact, Optimisable Representation for Dense Visual SLAM" and "DeepFactors: Real-Time Probabilistic Dense Monocular SLAM".

Notice

This repo have replicate the Jacobian calculation though network forward, But I'm not sure if the original author solved it this way, but I can achieve about the same speed as the paper with this approach. The method is that propagate the linear weight as the fellow fig:

Install

Create a conda environment and install requirements

conda create -n tandem python=3.7
conda activate tandem
conda install pytorch==1.5.1 torchvision==0.6.1 cudatoolkit=10.2 -c pytorch

pip install -r requirements.txt

Comment: The environment use PyTorch 1.5.1 and PyTorch-Lightning 0.7.6 because this was the environment we used for development.

Training

Config values are documented in config/default.yaml. You can start a training with

python train.py --config config/default.yaml path/to/out/folder DATA.ROOT_DIR $TANDEM_DATA_DIR

Result

The boundary is blur, the VAE encode to low dimension will lose high frequency.

Acknowledgements

Thanks for TANDEM, my code is build on it's framework.