MORE | Spring 2022

Monocular 3D Object Detection for Traffic Analysis

Recognizing and localizing objects in the 3D space is crucial for a more accurate representation of the environment for various use cases. While significant progress has been achieved with expensive LIDAR systems, 3D object detection is a challenging task given only a single image (without depth information). The research aims to implement a deep learning network that predicts 3D bounding boxes from Monocular images. The system will integrate into a resource-constrained traffic surveillance camera to solve tasks, such as road safety evaluation, trajectory estimation, object speed calculation, data archiving, 3D scene reconstructions, etc.