Research map / Label-efficient and robust learning / Recognition and long-tail
PI-H2T: Enhancing Long-Tailed Visual Recognition With Permutation-Invariant and Head-to-Tail Feature Fusion
Mengke Li, Zhikai Hu, Yang Lu, Weichao Lan, Yiu‐ming Cheung, Hui Fei Huang
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| add permutation-invariant feature fusion (PIF) and H2TF in place of H2T | DR · CIFAR100-LT (imbalance 100) | top-1 accuracy: 47.73 -> 49.22 | Improved |
| add PI-H2T in place of H2T | GCL · iNaturalist 2018 | top-1 accuracy: 71.62 -> 75.02 | Improved |
| add PI-H2T in place of H2T | RIDE · iNaturalist 2018 | top-1 accuracy: 73.11 -> 76.03 | Improved |
| add PI-H2T in place of H2T | DR · iNaturalist 2018 | top-1 accuracy: 71.81 -> 71.34 | Hurt |
| add PI-H2T in place of H2T | MisLAS · Places-LT | top-1 accuracy: 41.03 -> 40.69 | Hurt |
| add PI-H2T in place of H2T | GCL · Places-LT | top-1 accuracy: 40.73 -> 40.56 | Hurt |
| add PI-H2T to ViT-based LPT | LPT · Places-LT | top-1 accuracy: 49.70 -> 50.12 | Improved |
| add PI-H2T to ViT-based LIFT | LIFT · iNaturalist 2018 / ImageNet-LT | top-1 accuracy: 77.00 -> 78.16 on ImageNet-LT | Improved |
| use permutation-invariant fusion (PIF) vs non-PI baseline, Max operation (ablation) | H2TF stage 2 · CIFAR100-LT (imbalance 100) | top-1 accuracy: 45.68 -> 49.53 | Improved |
| use mean instead of Min as PI operation (ablation) | PIF · CIFAR100-LT (imbalance 100) | stage 2 top-1 accuracy: 48.53 -> 49.22 | Improved |
| add PIF in place of plain Conv1D (ablation) | ResNet-32 baseline · CIFAR100-LT (imbalance 100) | top-1 accuracy: 39.78 -> 42.19 | Improved |
| add PIF in place of SE block (ablation) | ResNet-32 baseline · CIFAR100-LT (imbalance 100) | top-1 accuracy: 40.55 -> 42.19 | Improved |
Closest papers by meaning
- Feature Fusion from Head to Tail for Long-Tailed Visual Recognition (AAAI 2024 · similarity 0.95)
- Deep Long-Tailed Learning: A Survey (TPAMI 2023 · similarity 0.90)
- Decoupling Representation and Classifier for Long-Tailed Recognition (ICLR 2020 · similarity 0.89)
- A Survey on Long-Tailed Visual Recognition (IJCV 2022 · similarity 0.89)
- Bag of Tricks for Long-Tailed Visual Recognition with Deep Convolutional Neural Networks (AAAI 2021 · similarity 0.89)
- Long-Tailed Recognition via Information-Preservable Two-Stage Learning (NeurIPS 2025 · similarity 0.89)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.
