Raghuram-Veeramallu/DiffTransBEV

BEV Representation of an Autonomous car using 6 RGB cameras by making use of Stable Diffusion Transformers

DiffTransBEV

Generating optimized noise-free BEV representaiton of Autonomous Vehicles (Self-driving cars) using Stable Diffusion Transformers (DiTs), Diffusion Models, SwinV2 Transformers and Lift-Splat-Shoot. NuScenes dataset was used for training and validation, while testing environment was simulated on CARLA simulator.

Installation

This project works best on:

• Python 3.7

Steps to recreate the environment

1. Create conda environment for the project

conda create --name av-bev python=3.7
conda activate av-bev

2. Install PyTorch

conda install pytorch torchvision -c pytorch