Monocular 3D Object Position Estimation with VLMs for Human-Robot Interaction
This work addresses the challenge of enabling robots to interact with objects based on visual and language inputs, though it is incremental as it adapts existing VLMs for a specific 3D task.
The paper tackled the problem of estimating 3D object positions from monocular RGB images using Vision-Language Models (VLMs) for human-robot interaction, achieving a median MAE of 13 mm and a five-fold improvement over a baseline, with about 25% of predictions being acceptable for robot interaction.
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.