Monocular 3D Object Position Estimation with VLMs for Human-Robot Interaction

TL;DR

This paper leverages pre-trained Vision-Language Models to estimate 3D object positions from monocular images for improved human-robot interaction, demonstrating significant accuracy improvements through fine-tuning.

Contribution

It introduces a novel approach to adapt VLMs for 3D coordinate detection in robotic contexts using a large dataset and specialized fine-tuning techniques.

Findings

Median MAE of 13 mm on test set

Five-fold improvement over baseline without fine-tuning

25% of predictions within acceptable interaction range

Abstract

Pre-trained general-purpose Vision-Language Models (VLM) hold the potential to enhance intuitive human-machine interactions due to their rich world knowledge and 2D object detection capabilities. However, VLMs for 3D coordinates detection tasks are rare. In this work, we investigate interactive abilities of VLMs by returning 3D object positions given a monocular RGB image from a wrist-mounted camera, natural language input, and robot states. We collected and curated a heterogeneous dataset of more than 100,000 images and finetuned a VLM using QLoRA with a custom regression head. By implementing conditional routing, our model maintains its ability to process general visual queries while adding specialized 3D position estimation capabilities. Our results demonstrate robust predictive performance with a median MAE of 13 mm on the test set and a five-fold improvement over a simpler baseline without finetuning. In about 25% of the cases, predictions are within a range considered acceptable for the robot to interact with objects.