# Perceiving Unseen 3D Objects by Poking the Objects

## TL;DR

This paper introduces a poking-based method enabling robots to autonomously discover, reconstruct, and recognize unseen 3D objects without prior models or extensive training data, improving robotic perception and manipulation.

## Contribution

The novel poking-based approach allows unsupervised discovery and high-quality reconstruction of unseen 3D objects for robotic perception.

## Key findings

- Successfully discovers unseen 3D objects in real-world data
- Reconstructs objects with high quality using unsupervised learning
- Facilitates robotic grasping and manipulation tasks

## Abstract

We present a novel approach to interactive 3D object perception for robots. Unlike previous perception algorithms that rely on known object models or a large amount of annotated training data, we propose a poking-based approach that automatically discovers and reconstructs 3D objects. The poking process not only enables the robot to discover unseen 3D objects but also produces multi-view observations for 3D reconstruction of the objects. The reconstructed objects are then memorized by neural networks with regular supervised learning and can be recognized in new test images. The experiments on real-world data show that our approach could unsupervisedly discover and reconstruct unseen 3D objects with high quality, and facilitate real-world applications such as robotic grasping. The code and supplementary materials are available at the project page: https://zju3dv.github.io/poking_perception.