AugmentedAutoencoder
A real-time RGB-based pipeline for object detection and 6D pose estimation using a denoising autoencoder trained on simulated 3D views.
What is AugmentedAutoencoder?
Augmented Autoencoders is a deep learning framework for 6D object pose estimation from RGB images. It uses a denoising autoencoder trained on synthetically rendered views of 3D models to predict object orientations without requiring real annotated data. The method generalizes across different sensors and inherently handles symmetries, providing a real-time pipeline for detection and pose estimation.
Computer vision researchers and engineers working on robotic perception, augmented reality, or industrial automation who need accurate 6D pose estimation from monocular RGB cameras.
It eliminates the need for expensive and time-consuming real-world pose annotation by training entirely on synthetic data via domain randomization. This makes it highly adaptable, robust to sensor variations, and capable of real-time performance.
Overview
Official Code: Implicit 3D Orientation Learning for 6D Object Detection from RGB Images
Use Cases
Best For
- Robotic bin-picking and manipulation tasks requiring precise object orientation
Related Projects
Found a gem we're missing?
Open-Awesome is built by the community, for the community. Submit a project, suggest an awesome list, or help improve the catalog on GitHub.
