mlsanityData to 14 Oct 2026

Coarse Is Better? A New Pipeline Towards Self-Supervised Learning with Uncurated Images

arXiv 2306.04244 ↗First posted 7 Jun 2023cs.CVcited by 7

Not on the map. Why?

Abstract (arXiv)

Most self-supervised learning (SSL) methods often work on curated datasets where the object-centric assumption holds. This assumption breaks down in uncurated images. Existing scene image SSL methods try to find the two views from original scene images that are well matched or dense, which is both complex and computationally heavy. This paper proposes a conceptually different pipeline: first find regions that are coarse objects (with adequate objectness), crop them out as pseudo object-centric images, then any SSL method can be directly applied as in a real object-centric dataset. That is, coarse crops benefits scene images SSL. A novel cropping strategy that produces coarse object box is proposed. The new pipeline and cropping strategy successfully learn quality features from uncurated datasets without ImageNet. Experiments show that our pipeline outperforms existing SSL methods (MoCo-v2, DenseCL and MAE) on classification, detection and segmentation tasks. We further conduct extensively ablations to verify that: 1) the pipeline do not rely on pretrained models; 2) the cropping strategy is better than existing object discovery methods; 3) our method is not sensitive to hyperparameters and data augmentations.

Key facts. Cited by 7 papers in the map (data through 14 Oct 2026). Builds on 43 of its 57 references that are in the map. First posted on arXiv 7 Jun 2023.

Builds on

43 of 57 references.

References · 43

  1. 2015Deep Residual Learning for Image Recognitioncited by 52,096
  2. 2020An Image is Worth 16x16 Words: Transformers for Image Recognition at Scalecited by 30,186
  3. 2019PyTorch: An Imperative Style, High-Performance Deep Learning Librarycited by 20,305
  4. 2014Microsoft COCO: Common Objects in Contextcited by 17,608
  5. 2020A Simple Framework for Contrastive Learning of Visual Representationscited by 12,049
  6. 2015Faster R-CNN: Towards Real-Time Object Detection with Region Proposal Networkscited by 9,639
  7. 2014ImageNet Large Scale Visual Recognition Challengecited by 8,498
  8. 2018Representation Learning with Contrastive Predictive Codingcited by 7,777
  9. 2021Masked Autoencoders Are Scalable Vision Learnerscited by 7,403
  10. 2017Mask R-CNNcited by 7,277
  11. 2020Bootstrap your own latent: A new approach to self-supervised Learningcited by 4,750
  12. 2016The Cityscapes Dataset for Semantic Urban Scene Understandingcited by 4,444
31 more
  1. 2020Unsupervised Learning of Visual Features by Contrasting Cluster Assignmentscited by 2,857
  2. 2016Pyramid Scene Parsing Networkcited by 2,479
  3. 2020Improved Baselines with Momentum Contrastive Learningcited by 2,287
  4. 2013Describing Textures in the Wildcited by 2,125
  5. 2013Fine-Grained Visual Classification of Aircraftcited by 1,726
  6. 2017Cascade R-CNN: Delving into High Quality Object Detectioncited by 1,276
  7. 2020What Makes for Good Views for Contrastive Learning?cited by 846
  8. 2022Exploring Plain Vision Transformer Backbones for Object Detectioncited by 724
  9. 2019Data-Efficient Image Recognition with Contrastive Predictive Codingcited by 673
  10. 2020Prototypical Contrastive Learning of Unsupervised Representationscited by 640
  11. 2018The Open Images Dataset V4: Unified image classification, object detection, and visual relationship detection at scalecited by 544
  12. 2020Dense Contrastive Learning for Self-Supervised Visual Pre-Trainingcited by 512
  13. 2018Rethinking ImageNet Pre-trainingcited by 377
  14. 2020Propagate Yourself: Exploring Pixel-Level Consistency for Unsupervised Visual Representation Learningcited by 297
  15. 2021DetCo: Unsupervised Contrastive Learning for Object Detectioncited by 205
  16. 2021Self-supervised Pretraining of Visual Features in the Wildcited by 198
  17. 2021Efficient Visual Pretraining with Contrastive Detectioncited by 150
  18. 2021Aligning Pretraining for Detection via Object-Level Contrastive Learningcited by 110
  19. 2020Bootstrap Latent-Predictive Representations for Multitask Reinforcement Learningcited by 102
  20. 2021Region Similarity Representation Learningcited by 96
  21. 2021Unsupervised Object-Level Representation Learning from Scene Imagescited by 72
  22. 2020Self-EMD: Self-Supervised Object Detection without ImageNetcited by 70
  23. 2021Revisiting Contrastive Methods for Unsupervised Learning of Visual Representationscited by 48
  24. 2016Selective Convolutional Descriptor Aggregation for Fine-Grained Image Retrievalcited by 44
  25. 2022Point-Level Region Contrast for Object Detection Pre-Trainingcited by 41
  26. 2021Self-Supervised Visual Representations Learning by Contrastive Mask Predictioncited by 31
  27. 2022Learning Where to Learn in Cross-View Self-Supervised Learningcited by 31
  28. 2021Object-Aware Cropping for Self-Supervised Learningcited by 22
  29. 2022Jigsaw-ViT: Learning Jigsaw Puzzles in Vision Transformercited by 21
  30. 2021A Random CNN Sees Objects: One Inductive Bias of CNN and Its Applicationscited by 15
  31. 2021Contrastive Attention Network with Dense Field Estimation for Face Completioncited by 6

Built on by

Cited by year (2026 to 14 Oct)

Citing papers per year: 2023: 1; 2024: 3; 2025: 2; 2026: 1.

7 papers.

  1. 2024Self-Supervised Visual Preference Alignmentcited by 31
  2. 2023Multi-Label Self-Supervised Learning with Scene Imagescited by 17
  3. 2024All You Need in Knowledge Distillation Is a Tailored Coordinate Systemcited by 6
  4. 2024Rectify the Regression Bias in Long-Tailed Object Detectioncited by 5
  5. 2025SHAPE : Self-Improved Visual Preference Alignment by Iteratively Generating Holistic Winnercited by 2
  6. 2025Beyond Instance Consistency: Investigating View Diversity in Self-supervised Learningcited by 1
  7. 2026Where to Look Matters: Rethinking Sub-Volume Sampling in 3D Medical Self-supervised Learningcited by 0