This is the project page for Finding Label and Model Errors in Perception Data With Learned Observation Assertions.
Please read the paper for full technical details.
In the root directory, run
pip install pandas
pip install -U git+https://github.com/lyft/nuscenes-devkit # This must be done from source
pip install pyclipper
pip install -e .
We provide an example of the Lyft Level 5 percetion dataset. We have provided model predictions for convenience, but you will need to download the dataset here.
All of the scripts are available in examples/lyft_level5
. In order to run the scripts, do the following:
- Set the data directories in
constants.py
. - Learn the priors with
learn_priors.py
. - Run LOA with
prior_lyft.py
.
You can visualize the results with viz_track.py
.
If you find this project useful, please cite us at
@article{kang2021finding,
title={Finding Label and Model Errors in Perception Data With Learned Observation Assertions},
author={Kang, Daniel and Arechiga, Nikos and Pillai, Sudeep and Bailis, Peter and Zaharia, Matei},
}
and contact us if you deploy LOA!