Skip to content

OFPNet is a baseline solution for Waymo Occupancy and Flow Prediction

License

Notifications You must be signed in to change notification settings

YoushaaMurhij/OFPNet

Repository files navigation

OFPNet

workflow License
OFPNet (Occupancy and flow predictive network) is developed for end-to-end prediction of occupancy map and flow using reccurent blocks, additional convolutional heads, etc.

OFPNet is a baseline solution for Waymo Occupancy and Flow Prediction

complete_scene observed_occupancy_rgb occluded_occupancy_rgb flow_rgb

Main Metrics

Metrics Observed Occupancy Occluded Occupancy Flow Flow-Grounded Occupancy
Model AUC Soft IoU AUC Soft IoU EPE AUC Soft IoU
UNet_LSTM 0.6559 0.4007 0.1227 0.0261 20.5876 0.5768 0.4280
UNet_LSTM_Head 0.6517 0.3859 0.1199 0.0225 20.1838 0.5840 0.4119
unext 0.6485 0.3580 0.0376 0.0084 21.6873 0.5598 0.4098
unext_head 0.7119 0.4257 0.1451 0.0309 21.6873 0.5691 0.4243

Basic Installation

Docker container:

Using nvidia-docker with cuda-11.3, Pytorch

cd path/to/workspace
git clone https://github.com/YoushaaMurhij/Occ_Flow_Pred.git
cd Occ_Flow_Pred/docker
./build.sh
cd ..
./docker/start.sh
./docker/into.sh

Conda environment (not recommended):

conda create --name occ_flow 
conda activate occ_flow
conda install pytorch torchvision cudatoolkit=11.3 -c pytorch
git clone https://github.com/YoushaaMurhij/Occ_Flow_Pred.git
cd Occ_Flow_Pred
pip install -r requirements.txt

# add Occ_Flow_Pred to PYTHONPATH by adding the following line to ~/.bashrc (change the path accordingly)
export PYTHONPATH="${PYTHONPATH}:/path/to/Occ_Flow_Pred/"

TODOs:

  • change data input
  • add more aux losses

Contribution:

Questions, suggestions and pull-requests are welcome!
Feel free to open an issue or a pull-request ☺️

Contacts:

Youshaa Murhij 📬 yosha[dot]morheg[at]phystech[dot]edu
Dmitry Yudin 📬 yudin[dot]da[at]mipt[dot]ru

About

OFPNet is a baseline solution for Waymo Occupancy and Flow Prediction

Topics

Resources

License

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published

Languages