# ROI-10D: Monocular Lifting of 2D Detection to 6D Pose and Metric Shape

## TL;DR

This paper introduces ROI-10D, a deep learning approach that lifts 2D detections into 3D space for accurate monocular 6D pose estimation and shape retrieval, achieving state-of-the-art results on KITTI3D.

## Contribution

The method uniquely integrates 2D detection, orientation, and scale into a 3D framework, improving metric accuracy and enabling synthetic data augmentation.

## Key findings

- Doubles AP on 3D pose metrics on KITTI3D
- Achieves state-of-the-art monocular 6D pose estimation
- Enables 3D shape recovery and synthetic data augmentation

## Abstract

We present a deep learning method for end-to-end monocular 3D object detection and metric shape retrieval. We propose a novel loss formulation by lifting 2D detection, orientation, and scale estimation into 3D space. Instead of optimizing these quantities separately, the 3D instantiation allows to properly measure the metric misalignment of boxes. We experimentally show that our 10D lifting of sparse 2D Regions of Interests (RoIs) achieves great results both for 6D pose and recovery of the textured metric geometry of instances. This further enables 3D synthetic data augmentation via inpainting recovered meshes directly onto the 2D scenes. We evaluate on KITTI3D against other strong monocular methods and demonstrate that our approach doubles the AP on the 3D pose metrics on the official test set, defining the new state of the art.