mlsanityData to 14 Oct 2026

Universal representations:The missing link between faces, text, planktons, and cat breeds

arXiv 1701.07275 ↗First posted 25 Jan 2017cs.CV stat.MLcited by 45

Not on the map. Why?

Abstract (arXiv)

With the advent of large labelled datasets and high-capacity models, the performance of machine vision systems has been improving rapidly. However, the technology has still major limitations, starting from the fact that different vision problems are still solved by different models, trained from scratch or fine-tuned on the target data. The human visual system, in stark contrast, learns a universal representation for vision in the early life of an individual. This representation works well for an enormous variety of vision problems, with little or no change, with the major advantage of requiring little training data to solve any of them. In this paper we investigate whether neural networks may work as universal representations by studying their capacity in relation to the “size” of a large combination of vision problems. We do so by showing that a single neural network can learn simultaneously several very different visual domains (from sketches to planktons and MNIST digits) as well as, or better than, a number of specialized networks. However, we also show that this requires to carefully normalize the information in the network, by using domain-specific scaling factors or, more generically, by using an instance normalization layer.

Key facts. Cited by 45 papers in the map (data through 14 Oct 2026). First posted on arXiv 25 Jan 2017.

Builds on

No reference list yet.

Built on by

Cited by year

Citing papers per year: 2021: 12; 2022: 14; 2023: 11; 2024: 5; 2025: 3.

45 papers.

Sort:

By area:

  1. 2022A Comprehensive Survey of Few-shot Learning: Evolution, Applications, Challenges, and Opportunitiescited by 163
  2. 2021MSeg: A Composite Dataset for Multi-domain Semantic Segmentationcited by 122
  3. 2022Head2Toe: Utilizing Intermediate Representations for Better Transfer Learningcited by 83
  4. 2023Modular Deep Learningcited by 83
  5. 2021Cross-domain Few-shot Learning with Task-specific Adapterscited by 80
  6. 2021Learning a Universal Template for Few-shot Dataset Generalizationcited by 61
  7. 2021Universal Representation Learning from Multiple Domains for Few-shot Classificationcited by 56
  8. 2021Style Normalization and Restitution for Domain Generalization and Adaptationcited by 46
  9. 2023Recent Advances on Federated Learning: A Systematic Surveycited by 41
  10. 2021INTERN: A New Learning Paradigm Towards General Visioncited by 33
  11. 2022Universal Representations: A Unified Look at Multiple Task and Domain Learningcited by 23
  12. 2022On the Domain Adaptation and Generalization of Pretrained Language Models: A Surveycited by 23
33 more
  1. 2021Text is Text, No Matter What: Unifying Text Recognition using Knowledge Distillationcited by 17
  2. 2022Learning useful representations for shifting tasks and distributionscited by 14
  3. 2023On the Effectiveness of LayerNorm Tuning for Continual Learning in Vision Transformerscited by 14
  4. 2023The Expressive Power of Tuning Only the Normalization Layerscited by 13
  5. 2024CDAD-Net: Bridging Domain Gaps in Generalized Category Discoverycited by 12
  6. 2023Learning to Drive Anywherecited by 10
  7. 2024UDON: Universal Dynamic Online distillatioN for generic image representationscited by 8
  8. 2024Discriminative Sample-Guided and Parameter-Efficient Feature Space Adaptation for Cross-Domain Few-Shot Learningcited by 7
  9. 2021Supervised Momentum Contrastive Learning for Few-Shot Classificationcited by 6
  10. 2021Rethinking Hard-Parameter Sharing in Multi-Domain Learningcited by 6
  11. 2022X-Learner: Learning Cross Sources and Tasks for Universal Visual Representationcited by 6
  12. 2022Hospital-Agnostic Image Representation Learning in Digital Pathologycited by 6
  13. 2022Could Giant Pretrained Image Models Extract Universal Representations?cited by 6
  14. 2021Boosting Binary Masks for Multi-Domain Learning through Affine Transformationscited by 5
  15. 2021Multi-Task, Multi-Domain Deep Segmentation with Shared Representations and Contrastive Regularization for Sparse Pediatric Datasetscited by 5
  16. 2022Generalizable multi-task, multi-domain deep segmentation of sparse pediatric imaging datasets via multi-scale contrastive regularization and multi-joint anatomical priorscited by 4
  17. 2023Visually Grounded Continual Language Learning with Selective Specializationcited by 4
  18. 2021Disentangling Transfer and Interference in Multi-Domain Learningcited by 3
  19. 2023Leveraging Normalization Layer in Adapters With Progressive Learning and Adaptive Distillation for Cross-Domain Few-Shot Learningcited by 3
  20. 2024Fine-tuning with Very Large Dropoutcited by 3
  21. 2025Revisit the Imbalance Optimization in Multi-task Learning: An Experimental Analysiscited by 3
  22. 2022An Empirical Study on Multi-Domain Robust Semantic Segmentationcited by 2
  23. 2023SGDA: Towards 3D Universal Pulmonary Nodule Detection via Slice Grouped Domain Attentioncited by 2
  24. 2022Seeking Common Ground While Reserving Differences: Multiple Anatomy Collaborative Framework for Undersampled MRI Reconstructioncited by 1
  25. 2022GPPF: A General Perception Pre-training Framework via Sparsely Activated Multi-Task Learningcited by 1
  26. 2023Multi-Domain Learning with Modulation Adapterscited by 1
  27. 2023Multi-domain improves out-of-distribution and data-limited scenarios for medical image analysiscited by 1
  28. 2025Federated Learning via Meta-Variational Dropoutcited by 1
  29. 2025Simplifying Multi-Task Architectures Through Task-Specific Normalizationcited by 1
  30. 2022E2-AEN: End-to-End Incremental Learning with Adaptively Expandable Networkcited by 0
  31. 2022MT-SNN: Spiking Neural Network that Enables Single-Tasking of Multiple Taskscited by 0
  32. 2023Cascaded Multi-task Adaptive Learning Based on Neural Architecture Searchcited by 0
  33. 2024Supervised Batch Normalizationcited by 0