https://github.com/stepankonev/waymo-motion-prediction-challenge-2022-multipath-plus-plus

Solution for Waymo Motion Prediction Challenge 2022. Our implementation of MultiPath++
https://github.com/stepankonev/waymo-motion-prediction-challenge-2022-multipath-plus-plus

autonomous-driving autonomous-vehicles cvpr2022 motion-prediction multipath pytorch sdc self-driving-car trajectory-prediction waymo-open-dataset workshop-autonomous-driving

Last synced: about 2 months ago
JSON representation

Solution for Waymo Motion Prediction Challenge 2022. Our implementation of MultiPath++

• Host: GitHub
• URL: https://github.com/stepankonev/waymo-motion-prediction-challenge-2022-multipath-plus-plus
• Owner: stepankonev
• License: other
• Created: 2022-06-05T14:42:52.000Z (over 2 years ago)
• Default Branch: main
• Last Pushed: 2023-04-22T20:05:00.000Z (over 1 year ago)
• Last Synced: 2024-08-01T03:42:53.116Z (5 months ago)
• Topics: autonomous-driving, autonomous-vehicles, cvpr2022, motion-prediction, multipath, pytorch, sdc, self-driving-car, trajectory-prediction, waymo-open-dataset, workshop-autonomous-driving
• Language: Python
• Homepage: https://arxiv.org/abs/2206.10041
• Size: 822 KB
• Stars: 364
• Watchers: 7
• Forks: 74
• Open Issues: 9
• Metadata Files:
  • Readme: README.md
  • License: LICENSE

Awesome Lists containing this project

README

        
# Waymo motion prediction challenge 2022: 3rd place solution (May, 26)

## Our implementation of [MultiPath++](https://arxiv.org/abs/2111.14973)

![](docs/architecture.png)

## General Info:

- 🏎️[**CVPR2022 Workshop on Autonomous Driving website**](https://cvpr2022.wad.vision)

- 📜[**Technical report**](https://arxiv.org/abs/2206.10041)   

- 🥉[**Waymo Motion Prediction Challenge Website**](https://waymo.com/open/challenges/2022/motion-prediction/)

- [❗Refactored code for our prize-winnig solution for Waymo Motion Prediction Challenge 2021](https://github.com/stepankonev/MotionCNN-Waymo-Open-Motion-Dataset)

## Team behind this solution:

Stepan Konev 

- [[LinkedIn]](https://www.linkedin.com/in/stepan-konev/)

- [[Twitter]](https://twitter.com/konevsteven)

- [[Facebook]](https://www.facebook.com/stepan.konev.31)

## Code Usage:

First we need to prepare data for training. The prerender script will convert the original data format into set of ```.npz``` files each containing the data for a single target agent. From ```code``` folder run

```

python3 prerender/prerender.py \

    --data-path /path/to/original/data \

    --output-path /output/path/to/prerendered/data \

    --n-jobs 24 \

    --n-shards 1 \

    --shard-id 0 \

    --config configs/prerender.yaml

```

Rendering is a memory consuming procedure so you may want to use ```n-shards > 1``` and running the script a few times using consecutive ```shard-id``` values

Once we have our data prepared we can run the training.

```

python3 train.py configs/final_RoP_Cov_Single.yaml

```

If you find this work interesting please ⭐️star and share this repo.

## Citation

If you find this work useful please cite us

```

@misc{https://doi.org/10.48550/arxiv.2206.10041,

  doi = {10.48550/ARXIV.2206.10041},

  url = {https://arxiv.org/abs/2206.10041},

  author = {Konev, Stepan},

  keywords = {Computer Vision and Pattern Recognition (cs.CV), FOS: Computer and information sciences, FOS: Computer and information sciences},

  title = {MPA: MultiPath++ Based Architecture for Motion Prediction},

  publisher = {arXiv},

  year = {2022},

  copyright = {arXiv.org perpetual, non-exclusive license}

}

```