mlsanityData to 7 Oct 2026

Long-tailed Instance Segmentation using Gumbel Optimized Loss

arXiv 2207.10936 ↗First posted 22 Jul 2022cs.CVcited by 12

Not on the map. Why?

Abstract (arXiv)

Major advancements have been made in the field of object detection and segmentation recently. However, when it comes to rare categories, the state-of-the-art methods fail to detect them, resulting in a significant performance gap between rare and frequent categories. In this paper, we identify that Sigmoid or Softmax functions used in deep detectors are a major reason for low performance and are sub-optimal for long-tailed detection and segmentation. To address this, we develop a Gumbel Optimized Loss (GOL), for long-tailed detection and segmentation. It aligns with the Gumbel distribution of rare classes in imbalanced datasets, considering the fact that most classes in long-tailed detection have low expected probability. The proposed GOL significantly outperforms the best state-of-the-art method by 1.1% on AP , and boosts the overall segmentation by 9.0% and detection by 8.0%, particularly improving detection of rare classes by 20.3%, compared to Mask-RCNN, on LVIS dataset. Code available at: https://github.com/kostas1515/GOL

Key facts. Cited by 12 papers in the map (data through 7 Oct 2026). Builds on 34 of its 44 references that are in the map. First posted on arXiv 22 Jul 2022.

Builds on

34 of 44 references.

References · 34

  1. 2015Deep Residual Learning for Image Recognitioncited by 51,809
  2. 2014Microsoft COCO: Common Objects in Contextcited by 17,498
  3. 2017Mask R-CNNcited by 7,259
  4. 2017Squeeze-and-Excitation Networkscited by 4,763
  5. 2016Aggregated Residual Transformations for Deep Neural Networkscited by 2,540
  6. 2019MMDetection: Open MMLab Detection Toolbox and Benchmarkcited by 1,379
  7. 2019Class-Balanced Loss Based on Effective Number of Samplescited by 1,239
  8. 2019LVIS: A Dataset for Large Vocabulary Instance Segmentationcited by 1,215
  9. 2019Learning Imbalanced Datasets with Label-Distribution-Aware Margin Losscited by 987
  10. 2019Decoupling Representation and Classifier for Long-Tailed Recognitioncited by 806
  11. 2019Large-Scale Long-Tailed Recognition in an Open Worldcited by 704
  12. 2020Long-tail learning via logit adjustmentcited by 605
22 more
  1. 2018Exploring the Limits of Weakly Supervised Pretrainingcited by 604
  2. 2019Hybrid Task Cascade for Instance Segmentationcited by 479
  3. 2019Cascade R-CNN: High Quality Object Detection and Instance Segmentationcited by 428
  4. 2020Balanced Meta-Softmax for Long-Tailed Visual Recognitioncited by 397
  5. 2020Equalization Loss for Long-Tailed Object Recognitioncited by 285
  6. 2020Long-Tailed Classification by Keeping the Good and Removing the Bad Momentum Causal Effectcited by 256
  7. 2021Distribution Alignment: A Unified Framework for Long-tail Visual Recognitioncited by 198
  8. 2015Cost Sensitive Learning of Deep Feature Representations from Imbalanced Datacited by 167
  9. 2020Overcoming Classifier Imbalance for Long-tail Object Detection with Balanced Group Softmaxcited by 152
  10. 2020Seesaw Loss for Long-Tailed Instance Segmentationcited by 146
  11. 2021Probabilistic two-stage detectioncited by 130
  12. 2019Imbalance Problems in Object Detection: A Reviewcited by 118
  13. 2020The Devil is in Classification: A Simple Framework for Long-tail Object Detection and Instance Segmentationcited by 108
  14. 2020Equalization Loss v2: A New Gradient Balance Approach for Long-tailed Object Detectioncited by 99
  15. 2021Adaptive Class Suppression Loss for Long-Tail Object Detectioncited by 55
  16. 2019Adjusting Decision Boundary for Class Imbalanced Learningcited by 52
  17. 2021Exploring Classification Equilibrium in Long-Tailed Object Detectioncited by 51
  18. 2020Forest R-CNN: Large-Vocabulary Long-Tailed Object Detection and Instance Segmentationcited by 41
  19. 2021On Model Calibration for Long-Tailed Object Detection and Instance Segmentationcited by 39
  20. 2020Large-Scale Object Detection in the Wild from Imbalanced Multi-Labelscited by 35
  21. 2021DropLoss for Long-Tail Instance Segmentationcited by 32
  22. 2020Background Splitting: Finding Rare Classes in a Sea of Backgroundcited by 5

Built on by

Cited by year

Citing papers per year: 2022: 3; 2023: 3; 2024: 6.

12 papers.

  1. 2022Objaverse: A Universe of Annotated 3D Objectscited by 1,522
  2. 2024A Systematic Review on Long-Tailed Learningcited by 27
  3. 2023HiLo: Exploiting High Low Frequency Relations for Unbiased Panoptic Scene Graph Generationcited by 19
  4. 2024Sparse Concept Bottleneck Models: Gumbel Tricks in Contrastive Learningcited by 11
  5. 2023Instance-Aware Repeat Factor Sampling for Long-Tailed Object Detectioncited by 7
  6. 2022Inverse Image Frequency for Long-tailed Image Recognitioncited by 6
  7. 2024Adaptive Parametric Activation: Unifying and Generalising Activation Functions Across Taskscited by 6
  8. 2024Rectify the Regression Bias in Long-Tailed Object Detectioncited by 5
  9. 2024Fractal Calibration for long-tailed object detectioncited by 5
  10. 2023Balanced Representation Learning for Long-tailed Skeleton-based Action Recognitioncited by 3
  11. 2022A Benchmark of Long-tailed Instance Segmentation with Noisy Labelscited by 0
  12. 2024PAT: Pixel-wise Adaptive Training for Long-tailed Segmentationcited by 0