/LiDAR-3d-object-detector

A LiDAR-based 3D object (vehicle) detector

Primary LanguagePython

CSC490H1: Making Your Self-driving Car Perceive the World

This repository contains our implementation for Module 1 of CSC490H1:
Making Your Self-driving Car Perceive the World.

Overview

The detector is trained on 27 sequences of the PandaSet LiDAR dataset.
The video below demonstrates vehicle detections on 960 frames of 12 testing sequences (different from the training sequences), with the green boxes being the ground truth labels and red boxes being the detections:

test.mp4

Getting started

  1. Install Miniconda:

    curl 'https://repo.anaconda.com/miniconda/Miniconda3-latest-Linux-x86_64.sh' > Miniconda.sh
bash Miniconda.sh
rm Miniconda.sh

  2. Close and re-open your terminal session.

  3. Change directories (cd) to where you cloned this repository.

  4. Create a new conda environment:

    conda env create --file environment.yml

  5. Activate your new environment:

    conda activate csc490

  6. Download PandaSet. After submitting your request to download the dataset, you will receive an email from Scale AI with instructions to download PandaSet in three parts. Download Part 1 only. After you have downloaded pandaset_0.zip, unzip the dataset as follows:

    unzip pandaset_0.zip -d <your_path_to_dataset>