paper-with-me

홈 › Papers

Measuring Visual Generalization in Continuous Control from Pixels

2020-10-13 · Jake Grigsby, Yanjun Qi

Self-supervised learning and data augmentation have significantly reduced the performance gap between state and image-based reinforcement learning agents in continuous control tasks. However, it is still unclear whether current techniques can face a variety of visual conditions required by real-world environments. We propose a challenging benchmark that tests agents' visual generalization by adding graphical variety to existing continuous control domains. Our empirical analysis shows that current methods struggle to generalize across a diverse set of visual changes, and we examine the specific factors of variation that make these tasks difficult. We find that data augmentation techniques outperform self-supervised learning approaches and that more significant image transformations provide better visual generalization \footnote{The benchmark and our augmented actor-critic implementation are open-sourced @ https://github.com/QData/dmc_remastered)

📄 PDF Abstract BibTeX arXiv:2010.06740

Code (2)

QData/dmc_remastered 공식 구현
jakegrigsby/dmc_remastered 공식 구현

Tasks

continuous-controlContinuous ControlData AugmentationReinforcement Learning (RL)Self-Supervised Learning

Similar Papers 제목 키워드 기반

Look where you look! Saliency-guided Q-networks for generalization in visual Reinforcement Learning

2022-09-16 · David Bertoin, Adil Zouitine, Mehdi Zouitine, Emmanuel Rachelson

Deep reinforcement learning policies, despite their outstanding efficiency in simulated visual control tasks, have shown disappointing ability to generalize across disturbances in the input training images. Changes in im…

Deep Reinforcement Learningreinforcement-learningReinforcement LearningReinforcement Learning (RL)

Don't Touch What Matters: Task-Aware Lipschitz Data Augmentation for Visual Reinforcement Learning

2022-02-21 · Zhecheng Yuan, Guozheng Ma, Yao Mu, Bo Xia 외

One of the key challenges in visual Reinforcement Learning (RL) is to learn policies that can generalize to unseen environments. Recently, data augmentation techniques aiming at enhancing data diversity have demonstrated…

Data AugmentationDiversityReinforcement Learning (RL)

Mastering Visual Continuous Control: Improved Data-Augmented Reinforcement Learning

2021-07-20 · ICLR 2022 4 · Denis Yarats, Rob Fergus, Alessandro Lazaric, Lerrel Pinto

We present DrQ-v2, a model-free reinforcement learning (RL) algorithm for visual continuous control. DrQ-v2 builds on DrQ, an off-policy actor-critic approach that uses data augmentation to learn directly from pixels. We…

continuous-controlContinuous ControlData AugmentationGPU+4

Sim-to-real reinforcement learning applied to end-to-end vehicle control

2020-12-14 · András Kalapos, Csaba Gór, Róbert Moni, István Harmati

In this work, we study vision-based end-to-end reinforcement learning on vehicle control problems, such as lane following and collision avoidance. Our controller policy is able to control a small-scale robot to follow th…

Collision Avoidancereinforcement-learningReinforcement Learning (RL)Transfer Learning

Potential Contrast: Properties, Equivalences, and Generalization to Multiple Classes

2025-05-02 · Wallace Peaslee, Anna Breger, Carola-Bibiane Schönlieb

Potential contrast is typically used as an image quality measure and quantifies the maximal possible contrast between samples from two classes of pixels in an image after an arbitrary grayscale transformation. It has bee…