# Virtual Training for a Real Application: Accurate Object-Robot Relative Localization without Calibration

## TL;DR

This paper presents a method for precise object-robot localization in uncalibrated scenes using synthetic training data, eliminating the need for real images and enabling accurate manipulation in uncontrolled environments.

## Contribution

The authors introduce a calibration-free, CNN-based localization approach trained solely on synthetic data, achieving millimetric accuracy without real image training.

## Key findings

- Achieves millimetric localization accuracy in uncalibrated scenes
- Uses only synthetic data for training, avoiding costly real data collection
- Provides a new dataset with real robot images for evaluation

## Abstract

Localizing an object accurately with respect to a robot is a key step for autonomous robotic manipulation. In this work, we propose to tackle this task knowing only 3D models of the robot and object in the particular case where the scene is viewed from uncalibrated cameras -- a situation which would be typical in an uncontrolled environment, e.g., on a construction site. We demonstrate that this localization can be performed very accurately, with millimetric errors, without using a single real image for training, a strong advantage since acquiring representative training data is a long and expensive process. Our approach relies on a classification Convolutional Neural Network (CNN) trained using hundreds of thousands of synthetically rendered scenes with randomized parameters. To evaluate our approach quantitatively and make it comparable to alternative approaches, we build a new rich dataset of real robot images with accurately localized blocks.