3D-BoundingBox
This project delivers 3D bounding box estimation using a PyTorch implementation that incorporates neural networks for precise orientation and dimensional prediction. Utilizing YOLOv3 for 2D detection, it calculates and overlays 3D locations on images. The project demands PyTorch, Cuda, and OpenCV, and provides pre-trained weights for quick setup. It allows batch processing of images and videos, demonstrating detection capabilities. Based on the Kitti dataset, it asserts strong performance. Future objectives include developing a customized YOLO network and implementing pose visualization, enhancing contributions to geometric conversion.