smallcorgi/3D-Deepbox
A TensorFlow implementation that predicts 3D bounding box size and orientation of objects from single 2D images using deep learning and geometric reasoning.
Not currently ranked — collecting fresh signals.
star history
This project implements the paper 3D Bounding Box Estimation Using Deep Learning and Geometry, using a MultiBin approach to jointly estimate 3D box dimensions and orientation from monocular images. It uses a VGG16-based architecture combined with geometric constraints to regress 3D bounding boxes for cars, pedestrians, and cyclists. The model is trained and evaluated on the KITTI autonomous driving dataset.
Frequently asked
- What is smallcorgi/3D-Deepbox?
- A TensorFlow implementation that predicts 3D bounding box size and orientation of objects from single 2D images using deep learning and geometric reasoning.
- Is 3D-Deepbox open source?
- Yes — smallcorgi/3D-Deepbox is open source, released under the MIT license.
- What language is 3D-Deepbox written in?
- smallcorgi/3D-Deepbox is primarily written in Python.
- How popular is 3D-Deepbox?
- smallcorgi/3D-Deepbox has 491 stars on GitHub.
- Where can I find 3D-Deepbox?
- smallcorgi/3D-Deepbox is on GitHub at https://github.com/smallcorgi/3D-Deepbox.