/motion-prediction-in-AVs

Exploring various methods to predict the motion of autonomous vehicles in presence of multiple agents

Primary LanguageJupyter Notebook

Motion Prediction in Autonomous Vehicles

Exploring various methods to predict the motion of autonomous vehicles in presence of multiple agents

This repository is a compilation of the various approaches we used in the Kaggle Competition:
Lyft Motion Prediction for Autonomous Vehicles
Goal of the competition: To built robust motion prediction models for self-driving vehicles.

Lyft has the largest publicly available dataset for this purpose: Dataset
A shortened Kaggle-friendly version of the same is available in the Data section on the main page.

If you were to attempt the competition problem statement yourself:

Following points might help you go ahead and make a baseline submission:

  1. To understand the dataset structure clearly, watch this: Intro
  2. Working with data can be difficult at the start. You might visit the L5Kit repo for sample codes and jupyter notebooks. I'd recommend you to try and run those notebooks yourself to get a decent idea about the dataset structure.