Learning Task-Oriented Grasping from Human Activity Datasets

TL;DR

This paper introduces a method for task-oriented robotic grasping that leverages human activity datasets to jointly estimate hand and object poses from RGB images, enabling robots to perform specific grasping tasks on novel objects.

Contribution

The paper presents a novel approach that jointly estimates hand and object poses from RGB images to improve task-oriented grasping accuracy, trained on human activity datasets.

Findings

Joint hand and object pose estimation improves accuracy.

Training with hand pose information enhances object pose prediction.

The method achieves state-of-the-art performance on real-world datasets.

Abstract

We propose to leverage a real-world, human activity RGB dataset to teach a robot Task-Oriented Grasping (TOG). We develop a model that takes as input an RGB image and outputs a hand pose and configuration as well as an object pose and a shape. We follow the insight that jointly estimating hand and object poses increases accuracy compared to estimating these quantities independently of each other. Given the trained model, we process an RGB dataset to automatically obtain the data to train a TOG model. This model takes as input an object point cloud and outputs a suitable region for task-specific grasping. Our ablation study shows that training an object pose predictor with the hand pose information (and vice versa) is better than training without this information. Furthermore, our results on a real-world dataset show the applicability and competitiveness of our method over state-of-the-art. Experiments with a robot demonstrate that our method can allow a robot to preform TOG on novel objects.