ShuoooooC/Navmamba_full

Object Goal Navigation with Recursive Implicit Maps

Method to Clone the repo

git clone --recurse-submodules https://huggingface.co/datasets/ShuoooooC/Navmamba_full onav_rim
cd onav_rim

Environment docker

docker pull shuochen0725/navmamba:2.10.0

# Create the container and use this 
conda activate navmamba

If not working, try this !!!!!!

Environment - 2, Install it locally

1. Environment Preaparation

conda remove -n navmamba --all
conda create -n navmamba python=3.10.13 cmake=3.14.0 
conda activate navmamba
git lfs install
git clone --recurse-submodules https://huggingface.co/datasets/ShuoooooC/Navmamba_full test
cd onav_rim

2. Install mamba and related package

# Mamba Install 
conda install pytorch==1.12.0 torchvision==0.13.0 torchaudio==0.12.0 cudatoolkit=11.6 -c pytorch -c nvidia -c conda-forge
conda install cudatoolkit-dev -c conda-forge

cd offline_bc/models/causal-conv1d
pip install -e .

cd ../mamba-1p1p1/
pip install -e .
cd ../../..

3. Install the package for the Habitat module

# Habitat install
cd dependencies/habitat-sim
pip install -r requirements.txt
apt-get update || true
# These are fairly ubiquitous packages and your system likely has them already,
# but if not, let's get the essentials for EGL support:
apt-get install -y --no-install-recommends \
     libjpeg-dev libglm-dev libgl1-mesa-glx libegl1-mesa-dev mesa-utils xorg-dev freeglut3-dev g++
./build.sh --headless --bullet  # for headless systems

# avoid logging
export GLOG_minloglevel=2
export MAGNUM_LOG=quiet
cd ../..

4. Install othe package that requires

# install CLIP
pip install ftfy regex tqdm
pip install git+https://github.com/openai/CLIP.git 
pip install openai-clip -i https://pypi.tuna.tsinghua.edu.cn/simple
# install the codebase
pip install -r requirements.txt
python setup.py develop --all

2. Unzip datasets

cd requires_datasets
unzip depth_fts.zip
unzip meta_infos_lmdb.zip
unzip objectnav_mp3d_70k.zip
unzip objectnav_mp3d.zip
unzip rgb_fts.zip
unzip sem_fts.zip
unzip scene_mp3d.zip
unzip test_assets.zip

# This will takes a lot of time, please be patience
git clone https://huggingface.co/sxyxs/mp3d_on
cd mp3d_on
mv rgb_fts/ code/onav_rim/data/datasets/objectnav/mp3d_70k_demos_prefts
cd zip_by_type
unzip depth_fts.zip 
# You need to move this folder in the code/onav_rim/data/datasets/objectnav/mp3d_70k_demos_prefts as well after unzip. Only need This folder "depth_fts".
# "Metion", onav_rim is the name for this project. Replace the correct path by yourself please
cd ../..

mv ddppo-models/ code/onav_rim/data/
mv rednet-models/ code/onav_rim/data/
mv code/onav_rim/data/ ./..
rm -rf code/
cd .. 

The data structure should look like this:

  ├── onav_rim/
  │  ├── data
  │  │  ├── ddppo-models/
  │  |  |  ├── gibson-2plus-resnet50.pth
  │  │  ├── rednet-models/
  │  │  │  ├── rednet_semmap_mp3d_tuned.pth
  │  │  ├── scene_datasets/
  │  │  │  ├── mp3d/
  │  │  ├── datasets
  │  │  │  ├── objectnav/
  │  │  │  │  ├── objectnav_mp3d_70k/
  │  │  │  │  │  ├── train/
  │  │  │  │  │── mp3d/
  │  │  │  │  │  ├── v1/
  │  │  │  │  │  │  ├── val/
  │  │  │  │  │── mp3d_70k_demos_prefits/
  │  │  │  │  │  ├── depth_fts/
  │  │  │  │  │  ├── meta_infos_lmdb/
  │  │  │  │  │  ├── rgb_fts/
  │  │  │  │  │  ├── sem_fts/  
  │  │  ├── test_assets/
  │  │  │  ├── objects/

Usage

1. Train and evaluate the model

Try this at first, if not work, do it manually

./job_scripts/three_tasks.sh   

For Leraning Rate Test

export PYTHONPATH=$PYTHONPATH:$(pwd)/dependencies/habitat-sim
export WORLD_SIZE=1
export MASTER_ADDR='gpu001'
export MASTER_PORT=10001

configfile=offline_bc/config/lr_large.yaml
CUDA_VISIBLE_DEVICES=0 python offline_bc/train_models.py --exp-config $configfile

export PYTHONPATH=$PYTHONPATH:$(pwd)/dependencies/habitat-sim
export WORLD_SIZE=1
export MASTER_ADDR='gpu002'
export MASTER_PORT=10002

configfile=offline_bc/config/lr_small.yaml
CUDA_VISIBLE_DEVICES=1 python offline_bc/train_models.py --exp-config $configfile

For Input Reshape

export PYTHONPATH=$PYTHONPATH:$(pwd)/dependencies/habitat-sim
export WORLD_SIZE=1
export MASTER_ADDR='gpu003'
export MASTER_PORT=10003

configfile=offline_bc/config/shape_model.yaml
CUDA_VISIBLE_DEVICES=2 python offline_bc/train_models.py --exp-config $configfile

We provide a trained model in [Dropbox](https://www.dropbox.com/scl/fo/bk0ok8ibxd5yirohqjmu6/h?rlkey=4riilnmjgk0e4o5rxt3prtsfl&dl=0).


## Acknowledgements
Some of the codes are built upon [habitat-imitation-baselines](https://github.com/Ram81/habitat-imitation-baselines/tree/master). Thanks them for their great work!


## Citation
If you find our work useful in your research, please consider citing:

@inproceedings{chen2023rim, author={Shizhe Chen and Thomas Chabal and Ivan Laptev and Cordelia Schmid}, title={Object Goal Navigation with Recursive Implicit Maps}, booktitle={Proc. of The International Conference on Intelligent Robots and Systems (IROS)}, year={2023} }