mlsanityData to 14 Oct 2026

Asymmetric Loss For Multi-Label Classification

arXiv 2009.14119 ↗First posted 29 Sep 2020cs.CV cs.LGcited by 381

Not on the map. Why?

Abstract (arXiv)

In a typical multi-label setting, a picture contains on average few positive labels, and many negative ones. This positive-negative imbalance dominates the optimization process, and can lead to under-emphasizing gradients from positive labels during training, resulting in poor accuracy. In this paper, we introduce a novel asymmetric loss ("ASL"), which operates differently on positive and negative samples. The loss enables to dynamically down-weights and hard-thresholds easy negative samples, while also discarding possibly mislabeled samples. We demonstrate how ASL can balance the probabilities of different samples, and how this balancing is translated to better mAP scores. With ASL, we reach state-of-the-art results on multiple popular multi-label datasets: MS-COCO, Pascal-VOC, NUS-WIDE and Open Images. We also demonstrate ASL applicability for other tasks, such as single-label classification and object detection. ASL is effective, easy to implement, and does not increase the training time or complexity. Implementation is available at: https://github.com/Alibaba-MIIL/ASL.

Key facts. Cited by 381 papers in the map (data through 14 Oct 2026). First posted on arXiv 29 Sep 2020.

Builds on

No reference list yet.

Built on by

Cited by year (2026 to 14 Oct)

Citing papers per year: 2021: 31; 2022: 62; 2023: 86; 2024: 69; 2025: 62; 2026: 71.

381 papers.

Sort:

By area:

  1. 2022AdaptFormer: Adapting Vision Transformers for Scalable Visual Recognitioncited by 719
  2. 2021ImageNet-21K Pretraining for the Massescited by 613
  3. 2023Recognize Anything: A Strong Image Tagging Modelcited by 309
  4. 2021Query2Label: A Simple Transformer Way to Multi-Label Classificationcited by 128
  5. 2021Re-labeling ImageNet: from Single to Multi-Labels, from Global to Localized Labelscited by 118
  6. 2022DualCoOp: Fast Adaptation to Multi-Label Recognition with Limited Annotationscited by 102
  7. 2023Detecting and Grounding Multi-Modal Media Manipulationcited by 99
  8. 2023Tag2Text: Guiding Vision-Language Model via Image Taggingcited by 86
  9. 2022Learning Multi-dimensional Edge Feature-based AU Relation Graph for Facial Action Unit Recognitioncited by 82
  10. 2021ML-Decoder: Scalable and Versatile Classification Headcited by 79
  11. 2022LPT: Long-tailed Prompt Tuning for Image Classificationcited by 69
  12. 2023Open-Set Image Tagging with Multi-Grained Text Supervisioncited by 67
38 more
  1. 2023Which Tokens to Use? Investigating Token Reduction in Vision Transformerscited by 60
  2. 2023Detecting and Grounding Multi-Modal Media Manipulation and Beyondcited by 56
  3. 2021Residual Attention: A Simple but Effective Method for Multi-Label Recognitioncited by 52
  4. 2022Texts as Images in Prompt Tuning for Multi-Label Image Recognitioncited by 51
  5. 2023Natural Language-Assisted Sign Language Recognitioncited by 46
  6. 2021Injecting Semantic Concepts into End-to-End Image Captioningcited by 41
  7. 2023CDUL: CLIP-Driven Unsupervised Learning for Multi-Label Image Classificationcited by 39
  8. 2022Missingness Bias in Model Debuggingcited by 38
  9. 2022Large Loss Matters in Weakly Supervised Multi-Label Classificationcited by 38
  10. 2025MAT-Agent: Adaptive Multi-Agent Training Optimizationcited by 35
  11. 2022Comprehending and Ordering Semantics for Image Captioningcited by 33
  12. 2023FoodLMM: A Versatile Food Assistant using Large Multi-modal Modelcited by 32
  13. 2021Simple and Robust Loss Design for Multi-Label Learning with Missing Labelscited by 30
  14. 2022Semantic-Aware Representation Blending for Multi-Label Image Recognition with Partial Labelscited by 29
  15. 2024AUFormer: Vision Transformers are Parameter-Efficient Facial Action Unit Detectorscited by 29
  16. 2022A Sketch Is Worth a Thousand Words: Image Retrieval with Text and Sketchcited by 28
  17. 2025Can Masked Autoencoders Also Listen to Birds?cited by 26
  18. 2021Multi-label Classification with Partial Annotations using Class-aware Selective Losscited by 25
  19. 2021Semantic Diversity Learning for Zero-Shot Multi-label Classificationcited by 24
  20. 2022Acknowledging the Unknown for Multi-label Learning with Single Positive Labelscited by 24
  21. 2021sigmoidF1: A Smooth F1 Score Surrogate Loss for Multilabel Classificationcited by 23
  22. 2023DualCoOp++: Fast and Effective Adaptation to Multi-Label Recognition with Limited Annotationscited by 22
  23. 2023Towards long-tailed, multi-label disease classification from chest X-ray: Overview of the CXR-LT challengecited by 22
  24. 2021MlTr: Multi-label Classification with Transformercited by 21
  25. 2023Category Query Learning for Human-Object Interaction Classificationcited by 21
  26. 2023PatchCT: Aligning Patch Set and Label Set with Conditional Transport for Multi-Label Image Classificationcited by 21
  27. 2023UniChest: Conquer-and-Divide Pre-training for Multi-Source Chest X-Ray Classificationcited by 21
  28. 2022Heterogeneous Semantic Transfer for Multi-label Recognition with Partial Labelscited by 19
  29. 2022Multi-Label Retinal Disease Classification using Transformerscited by 19
  30. 2023PARFormer: Transformer-based Multi-Task Network for Pedestrian Attribute Recognitioncited by 19
  31. 2023TagCLIP: A Local-to-Global Framework to Enhance Open-Vocabulary Multi-Label Classification of CLIP Without Trainingcited by 19
  32. 2021Multi-Label Image Classification with Contrastive Learningcited by 18
  33. 2022Don't Forget Me: Accurate Background Recovery for Text Removal via Modeling Local-Global Contextcited by 18
  34. 2023Multi-label affordance mapping from egocentric visioncited by 18
  35. 2021Sewer-ML: A Multi-Label Sewer Defect Classification Dataset and Benchmarkcited by 17
  36. 2022Spatial Consistency Loss for Training Multi-Label Classifiers from Single-Label Annotationscited by 17
  37. 2022PIDray: A Large-scale X-ray Benchmark for Real-World Prohibited Item Detectioncited by 17
  38. 2023Learning in Imperfect Environment: Multi-Label Classification with Long-Tailed Distribution and Partial Labelscited by 17

50 most cited of 381 All 381 (JSON)