Motion-Prediction-for-Autonomous-Vehicle

KO | EN

gitlite — search

#hacktoberfest #react #ai #typescript #rust #deep-learning #linux #docker #macos #windows #kubernetes #awesome-list #security #openai

Motion-Prediction-for-Autonomous-Vehicle

(★ 15)

agent car's trajectory prediction for autonomous vehicles using Lyft dataset

Motion-Prediction-for-Autonomous-Vehicle Latest Version Download

Download Latest Version (.zip)

// repository documentation

Was this content helpful?

★ 0 (0 ratings)

# Motion Prediction for Autonomous Vehicles

![frame_695299_pred](https://raw.githubusercontent.com/aaronzguan/Motion-Prediction-for-Autonomous-Vehicle/main/visualize_xception/frame_695299_pred.png)

This is the final project of CMU 16824 Visual Learning and Recognition. We designed a simple regression CNN pipeline to predict the agents' motion for the autonomous vehicle and trained on multi-channel images with negative log-likelihood loss function. The supported backbones include ResNet28, Xception41, and DenseNet121.

This repo heavily rely on the l5kit tool, please follow below steps to install dependencies and download the Lyft dataset first.

## Run Model

1. Install [Lyft Tool Kits](https://github.com/lyft/l5kit):

```
   git clone https://github.com/lyft/l5kit.git
   cd l5kit/l5kit
   ```

If you don't care about dependency builds or you're having troubles with packages resolution, you can modify the `setup.py` and then running:

```
   pip install -e ."[dev]"
   ```

2. Download Lyft dataset following instructions on [Kaggle](https://www.kaggle.com/c/lyft-motion-prediction-autonomous-vehicles/data) or simply using Kaggle API:

```
   kaggle competitions download -c lyft-motion-prediction-autonomous-vehicles
   ```

After unzip all files, you should get the folder structure same as this:

```
   ├── Lyft-data
   │   ├── README.md
   │   ├── aerial_map
   │   │   ├── LICENSE
   │   │   ├── aerial_map.png
   │   │   ├── feedback.txt
   │   │   └── nearmap_images
   │   ├── meta.json
   │   ├── multi_mode_sample_submission.csv
   │   ├── scenes
   │   │   ├── mask.npz
   │   │   ├── sample.zarr
   │   │   ├── test.zarr
   │   │   ├── train.zarr
   │   │   └── validate.zarr
   │   ├── semantic_map
   │   │   ├── LICENSE
   │   │   ├── feedback.txt
   │   │   └── semantic_map.pb
   │   └── single_mode_sample_submission.csv
   ```

3. Modify the arguments in `agent_motion_config.yaml`, and change the dataset path in the `train.py` then you are all set to go by `python train.py` !

## Extra

Final report will come soon!

Please raise issues or [contact me](mailto:aaron.z.guan@hotmail.com) if you need our trained models for ResNet28 and Xception41.

Motion-Prediction-for-Autonomous-Vehicle

Motion-Prediction-for-Autonomous-Vehicle Latest Version Download

Today's picks

Was this content helpful?

Recent Feedback

Motion-Prediction-for-Autonomous-Vehicle

Motion-Prediction-for-Autonomous-Vehicle Latest Version Download

Today's picks

Was this content helpful?

Recent Feedback

Download README