Unsupervised Learning of Object Keypoints for Perception and Control

TL;DR

This paper introduces Transporter, an unsupervised neural network architecture that learns geometric object keypoints from raw video frames, improving object tracking and enabling efficient control and exploration in reinforcement learning.

Contribution

The paper presents a novel unsupervised method for discovering object keypoints that enhance control and exploration in reinforcement learning tasks.

Findings

Keypoints enable highly sample-efficient reinforcement learning.

Long-term object tracking is more accurate than previous methods.

Controlling keypoints reduces exploration space significantly.

Abstract

The study of object representations in computer vision has primarily focused on developing representations that are useful for image classification, object detection, or semantic segmentation as downstream tasks. In this work we aim to learn object representations that are useful for control and reinforcement learning (RL). To this end, we introduce Transporter, a neural network architecture for discovering concise geometric object representations in terms of keypoints or image-space coordinates. Our method learns from raw video frames in a fully unsupervised manner, by transporting learnt image features between video frames using a keypoint bottleneck. The discovered keypoints track objects and object parts across long time-horizons more accurately than recent similar methods. Furthermore, consistent long-term tracking enables two notable results in control domains -- (1) using the keypoint co-ordinates and corresponding image features as inputs enables highly sample-efficient reinforcement learning; (2) learning to explore by controlling keypoint locations drastically reduces the search space, enabling deep exploration (leading to states unreachable through random action exploration) without any extrinsic rewards.