mlsanityData to 7 Oct 2026

Long-Tail Temporal Action Segmentation with Group-wise Temporal Logit Adjustment

arXiv 2408.09919 ↗First posted 19 Aug 2024cs.CVcited by 3

Not on the map. Why?

Abstract (arXiv)

Procedural activity videos often exhibit a long-tailed action distribution due to varying action frequencies and durations. However, state-of-the-art temporal action segmentation methods overlook the long tail and fail to recognize tail actions. Existing long-tail methods make class-independent assumptions and struggle to identify tail classes when applied to temporal segmentation frameworks. This work proposes a novel group-wise temporal logit adjustment~(G-TLA) framework that combines a group-wise softmax formulation while leveraging activity information and action ordering for logit adjustment. The proposed framework significantly improves in segmenting tail actions without any performance loss on head actions.

Key facts. Cited by 3 papers in the map (data through 7 Oct 2026). Builds on 33 of its 53 references that are in the map. First posted on arXiv 19 Aug 2024.

Builds on

33 of 53 references.

References · 33

  1. 2017Quo Vadis, Action Recognition? A New Model and the Kinetics Datasetcited by 3,259
  2. 2019Class-Balanced Loss Based on Effective Number of Samplescited by 1,239
  3. 2019Learning Imbalanced Datasets with Label-Distribution-Aware Margin Losscited by 987
  4. 2019Decoupling Representation and Classifier for Long-Tailed Recognitioncited by 806
  5. 2018Learning to Reweight Examples for Robust Deep Learningcited by 670
  6. 2020Long-tail learning via logit adjustmentcited by 605
  7. 2017A systematic study of the class imbalance problem in convolutional neural networkscited by 510
  8. 2016Temporal Convolutional Networks for Action Segmentation and Detectioncited by 504
  9. 2019BBN: Bilateral-Branch Network with Cumulative Learning for Long-Tailed Visual Recognitioncited by 403
  10. 2019MS-TCN: Multi-Stage Temporal Convolutional Network for Action Segmentationcited by 370
  11. 2020Equalization Loss for Long-Tailed Object Recognitioncited by 285
  12. 2022Assembly101: A Large-Scale Multi-View Video Dataset for Understanding Procedural Activitiescited by 284
21 more
  1. 2018What is the Effect of Importance Weighting in Deep Learning?cited by 250
  2. 2021ASFormer: Transformer for Action Segmentationcited by 166
  3. 2020Overcoming Classifier Imbalance for Long-tail Object Detection with Balanced Group Softmaxcited by 152
  4. 2020Seesaw Loss for Long-Tailed Instance Segmentationcited by 146
  5. 2015Unsupervised Learning from Narrated Instruction Videoscited by 116
  6. 2015Relay Backpropagation for Effective Learning of Deep Convolutional Neural Networkscited by 115
  7. 2020The Devil is in Classification: A Simple Framework for Long-tail Object Detection and Instance Segmentationcited by 108
  8. 2021ACE: Ally Complementary Experts for Solving Long-Tailed Recognition in One-Shotcited by 86
  9. 2022Temporal Action Segmentation: An Analysis of Modern Techniquescited by 85
  10. 2023Diffusion Action Segmentationcited by 74
  11. 2020Learning to Segment the Tailcited by 54
  12. 2019Adjusting Decision Boundary for Class Imbalanced Learningcited by 52
  13. 2021Global2Local: Efficient Structure Search for Video Action Segmentationcited by 42
  14. 2019A Hybrid RNN-HMM Approach for Weakly Supervised Temporal Action Segmentationcited by 36
  15. 2021VideoLT: Large-scale Long-tailed Video Recognitioncited by 35
  16. 2021Adaptive Logit Adjustment Loss for Long-Tailed Visual Recognitioncited by 33
  17. 2021Coarse to Fine Multi-Resolution Temporal Convolutional Networkcited by 28
  18. 2016Reviving Threshold-Moving: a Simple Plug-in Bagging Ensemble for Binary and Multiclass Imbalanced Datacited by 23
  19. 2023Use Your Head: Improving Long-Tail Video Recognitioncited by 22
  20. 2019To Balance or Not to Balance: A Simple-yet-Effective Approach for Learning with Long-Tailed Distributionscited by 20
  21. 2022Leveraging Action Affinity and Continuity for Semi-supervised Temporal Action Segmentationcited by 17

Built on by

Cited by year (2026 to 7 Oct)

Citing papers per year: 2025: 2; 2026: 1.

3 papers.

  1. 2025Context-Enhanced Memory-Refined Transformer for Online Action Detectioncited by 13
  2. 2025Distinguishing Visually Similar Actions: Prompt-Guided Semantic Prototype Modulation for Few-Shot Action Recognitioncited by 0
  3. 2026Seeing Semantic Shift: Difference-Aware Sentence-Level Temporal Segmentation of Sign Language Videoscited by 0