Where2Act: From Pixels to Actions for Articulated 3D Objects
This work is significant for robotic agents that need to interact with articulated objects by predicting specific action locations and types from visual data, contributing to more meaningful human-robot interaction.
This paper addresses the problem of identifying actionable information for interacting with articulated 3D objects from visual input. The proposed network predicts possible actions (e.g., pushing, pulling) and the regions of articulated parts that will move under force, enabling agents to interact with objects like opening a drawer by pulling its handle.
One of the fundamental goals of visual perception is to allow agents to meaningfully interact with their environment. In this paper, we take a step towards that long-term goal -- we extract highly localized actionable information related to elementary actions such as pushing or pulling for articulated objects with movable parts. For example, given a drawer, our network predicts that applying a pulling force on the handle opens the drawer. We propose, discuss, and evaluate novel network architectures that given image and depth data, predict the set of actions possible at each pixel, and the regions over articulated parts that are likely to move under the force. We propose a learning-from-interaction framework with an online data sampling strategy that allows us to train the network in simulation (SAPIEN) and generalizes across categories. Check the website for code and data release: https://cs.stanford.edu/~kaichun/where2act/