mlsanityData to 13 Oct 2026

On Pre-Training for Visuo-Motor Control: Revisiting a Learning-from-Scratch Baseline

arXiv 2212.05749 ↗First posted 12 Dec 2022cs.LG cs.CV cs.ROcited by 55

Not on the map. Why?

Abstract (arXiv)

In this paper, we examine the effectiveness of pre-training for visuo-motor control tasks. We revisit a simple Learning-from-Scratch (LfS) baseline that incorporates data augmentation and a shallow ConvNet, and find that this baseline is surprisingly competitive with recent approaches (PVR, MVP, R3M) that leverage frozen visual representations trained on large-scale vision datasets -- across a variety of algorithms, task domains, and metrics in simulation and on a real robot. Our results demonstrate that these methods are hindered by a significant domain gap between the pre-training datasets and current benchmarks for visuo-motor control, which is alleviated by finetuning. Based on our findings, we provide recommendations for future research in pre-training for control and hope that our simple yet strong baseline will aid in accurately benchmarking progress in this area.

Key facts. Cited by 55 papers in the map (data through 13 Oct 2026). Builds on 40 of its 58 references that are in the map. First posted on arXiv 12 Dec 2022.

Builds on

40 of 58 references.

References · 40

  1. 2015Deep Residual Learning for Image Recognitioncited by 51,950
  2. 2018BERT: Pre-training of Deep Bidirectional Transformers for Language Understandingcited by 40,304
  3. 2021Learning Transferable Visual Models From Natural Language Supervisioncited by 33,588
  4. 2020Language Models are Few-Shot Learnerscited by 32,320
  5. 2020An Image is Worth 16x16 Words: Transformers for Image Recognition at Scalecited by 30,043
  6. 2017Proximal Policy Optimization Algorithmscited by 14,592
  7. 2014ImageNet Large Scale Visual Recognition Challengecited by 8,476
  8. 2018Representation Learning with Contrastive Predictive Codingcited by 7,712
  9. 2019Momentum Contrast for Unsupervised Visual Representation Learningcited by 7,492
  10. 2021Masked Autoencoders Are Scalable Vision Learnerscited by 7,364
  11. 2022PaLM: Scaling Language Modeling with Pathwayscited by 5,317
  12. 2022Flamingo: a Visual Language Model for Few-Shot Learningcited by 4,363
28 more
  1. 2015Continuous control with deep reinforcement learningcited by 3,290
  2. 2021Ego4D: Around the World in 3,000 Hours of Egocentric Videocited by 1,535
  3. 2017Domain Randomization for Transferring Deep Neural Networks from Simulation to the Real Worldcited by 1,411
  4. 2015Unsupervised Visual Representation Learning by Context Predictioncited by 1,064
  5. 2018DeepMind Control Suitecited by 944
  6. 2020CURL: Contrastive Unsupervised Representations for Reinforcement Learningcited by 747
  7. 2017Learning Complex Dexterous Manipulation with Deep Reinforcement Learning and Demonstrationscited by 687
  8. 2022R3M: A Universal Visual Representation for Robot Manipulationcited by 683
  9. 2020Image Augmentation Is All You Need: Regularizing Deep Reinforcement Learning from Pixelscited by 576
  10. 2020Reinforcement Learning with Augmented Datacited by 456
  11. 2022Temporal Difference Learning for Model Predictive Controlcited by 384
  12. 2021Mastering Visual Continuous Control: Improved Data-Augmented Reinforcement Learningcited by 361
  13. 2020Data-Efficient Reinforcement Learning with Self-Predictive Representationscited by 311
  14. 2019Improving Sample Efficiency in Model-Free Reinforcement Learning from Imagescited by 268
  15. 2021When Vision Transformers Outperform ResNets without Pre-training or Strong Data Augmentationscited by 259
  16. 2022Masked Visual Pre-training for Motor Controlcited by 246
  17. 2017Asymmetric Actor Critic for Image-Based Robot Learningcited by 205
  18. 2022The Unsurprising Effectiveness of Pre-Trained Vision Models for Controlcited by 155
  19. 2020Generalization in Reinforcement Learning by Soft Data Augmentationcited by 134
  20. 2021Stabilizing Deep Q-Learning with ConvNets and Vision Transformers under Data Augmentationcited by 103
  21. 2021RRL: Resnet as representation for Reinforcement Learningcited by 89
  22. 2020Automatic Data Augmentation for Generalization in Deep Reinforcement Learningcited by 71
  23. 2016Unsupervised Perceptual Rewards for Imitation Learningcited by 64
  24. 2022VRL3: A Data-Driven Framework for Visual Deep Reinforcement Learningcited by 51
  25. 2022Visual Reinforcement Learning with Self-Supervised 3D Representationscited by 45
  26. 2022A Comprehensive Survey of Data Augmentation in Visual Reinforcement Learningcited by 24
  27. 2021Reinforcement Learning with Latent Flowcited by 18
  28. 2022On the Feasibility of Cross-Task Transfer with Model-Based Reinforcement Learningcited by 15

Built on by

Cited by year (2026 to 13 Oct)

Citing papers per year: 2023: 26; 2024: 19; 2025: 9; 2026: 1.

55 papers.

Sort:

By area:

  1. 2023MimicPlay: Long-Horizon Imitation Learning by Watching Human Playcited by 216
  2. 2023Where are we in the search for an Artificial Visual Cortex for Embodied Intelligence?cited by 216
  3. 2023LIV: Language-Image Representations and Rewards for Robotic Controlcited by 181
  4. 2023RoboAgent: Generalization and Efficiency in Robot Manipulation via Semantic Augmentations and Action Chunkingcited by 168
  5. 2024HumanPlus: Humanoid Shadowing and Imitation from Humanscited by 166
  6. 2023Look Before You Leap: Unveiling the Power of GPT-4V in Robotic Vision-Language Planningcited by 161
  7. 2024Gen2Act: Human Video Generation in Novel Scenarios enables Generalizable Robot Manipulationcited by 146
  8. 20243D Diffusion Policy: Generalizable Visuomotor Policy Learning via Simple 3D Representationscited by 115
  9. 2023Toward General-Purpose Robots via Foundation Models: A Survey and Meta-Analysiscited by 102
  10. 2023GNFactor: Multi-Task Real Robot Learning with Generalizable Neural Feature Fieldscited by 99
  11. 2024General Flow as Foundation Affordance for Scalable Robot Learningcited by 81
  12. 2023Learning Generalizable Manipulation Policies with Object-Centric 3D Representationscited by 60
38 more
  1. 2023Diffusion Reward: Learning Rewards via Conditional Video Diffusioncited by 56
  2. 2023Imitation Bootstrapped Reinforcement Learningcited by 55
  3. 2024Generalizable Humanoid Manipulation with 3D Diffusion Policiescited by 50
  4. 2023TAIL: Task-specific Adapters for Imitation Learning with Large Pretrained Modelscited by 48
  5. 2024Point Cloud Matters: Rethinking the Impact of Different Observation Spaces on Robot Learningcited by 45
  6. 2023An Unbiased Look at Datasets for Visuo-Motor Pre-Trainingcited by 42
  7. 2024Learning to Manipulate Anywhere: A Visual Generalizable Framework For Reinforcement Learningcited by 37
  8. 2024Learning Manipulation by Predicting Interactioncited by 36
  9. 2023Lossless Adaptation of Pretrained Vision Models For Robotic Manipulationcited by 28
  10. 2023Revisiting Plasticity in Visual Reinforcement Learning: Data, Modules and Training Stagescited by 27
  11. 2024Hierarchical World Models as Visual Whole-Body Humanoid Controllerscited by 25
  12. 2023SpawnNet: Learning Generalizable Visuomotor Skills from Pre-trained Networkscited by 24
  13. 2023Unleashing the Power of Pre-trained Language Models for Offline Reinforcement Learningcited by 24
  14. 2024Robots Pre-train Robots: Manipulation-Centric Robotic Representation from Large-Scale Robot Datasetscited by 23
  15. 2023For Pre-Trained Vision Models in Motor Control, Not All Policy Learning Methods are Created Equalcited by 22
  16. 2023Chain-of-Thought Predictive Controlcited by 18
  17. 2024SAM-E: Leveraging Visual Foundation Model with Sequence Imitation for Embodied Manipulationcited by 18
  18. 2024Semantically Controllable Augmentations for Generalizable Robot Learningcited by 17
  19. 2023MoVie: Visual Model-Based Policy Adaptation for View Generalizationcited by 16
  20. 2024Point Cloud Models Improve Visual Robustness in Robotic Learnerscited by 13
  21. 2024A Recipe for Unbounded Data Augmentation in Visual Reinforcement Learningcited by 13
  22. 2023Generalizable Visual Reinforcement Learning with Segment Anything Modelcited by 12
  23. 2025Learning Massively Multitask World Models for Continuous Controlcited by 11
  24. 2023Human-oriented Representation Learning for Robotic Manipulationcited by 10
  25. 2024Premier-TACO is a Few-Shot Policy Learner: Pretraining Multitask Representation via Temporal Action-Driven Contrastive Losscited by 10
  26. 2023Pre-training with Synthetic Data Helps Offline Reinforcement Learningcited by 9
  27. 2023Learning Better with Less: Effective Augmentation for Sample-Efficient Visual Reinforcement Learningcited by 8
  28. 2025Spatial-Temporal Aware Visuomotor Diffusion Policy Learningcited by 8
  29. 2024The Surprising Ineffectiveness of Pre-Trained Visual Representations for Model-Based Reinforcement Learningcited by 7
  30. 2025A Data-Centric Revisit of Pre-Trained Vision Models for Robot Learningcited by 7
  31. 2023FoMo Rewards: Can we cast foundation models as reward functions?cited by 4
  32. 2024Overcoming Knowledge Barriers: Online Imitation Learning from Visual Observation with Pretrained World Modelscited by 2
  33. 2025Attentive Feature Aggregation or: How Policies Learn to Stop Worrying about Robustness and Attend to Task-Relevant Visual Cuescited by 2
  34. 2023H-InDex: Visual Reinforcement Learning with Hand-Informed Representations for Dexterous Manipulationcited by 1
  35. 2023A Statistical Guarantee for Representation Transfer in Multitask Imitation Learningcited by 1
  36. 2024AdaDemo: Data-Efficient Demonstration Expansion for Generalist Robotic Agentcited by 1
  37. 2025Efficient Reinforcement Learning by Guiding World Models with Non-Curated Datacited by 1
  38. 2025Disentangled Object-Centric Image Representation for Robotic Manipulationcited by 1

50 most cited of 55 All 55 (JSON)