Bird’s Eye View Perception

This repository contains my work on Bird’s Eye View (BEV) transformation using both classical and deep learning methods.
It includes inverse perspective mapping and deep learning-based multi-camera fusion by implementing UNetXST from the paper "A Sim2Real Deep Learning Approach for the Transformation of Images from Multiple Vehicle-Mounted Cameras to a Semantically Segmented Image in Bird's Eye View."

Please download the dataset from here.

---

📌 Features

• BEV generation using Inverse Perspective Mapping (IPM)
• Multi-camera to BEV transformation using deep learning approaches
• Implementation of the UNetXST architecture for BEV semantic segmentation
• Training and evaluation on multiple front-facing camera inputs