Research map / Label-efficient and robust learning / Recognition and long-tail
Long-Tailed Visual Recognition via Self-Heterogeneous Integration With Knowledge Excavation
Read the paperCheck your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (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 SHIKE to the CIFAR100-LT recognition setup | ResNet-32 · CIFAR100-LT IF100 | accuracy: 54.2 -> 56.3 | Helped |
| Add SHIKE to the CIFAR100-LT recognition setup | ResNet-32 · CIFAR100-LT IF50 | accuracy: 58.2 -> 59.8 | Helped |
| Train SHIKE for 400 epochs instead of the standard ImageNet-LT schedule | SHIKE with ResNet-50 · ImageNet-LT | Top-1 accuracy: 59.7 -> 60.3 | Helped |
| Use SHIKE rather than NCL (3N) | ImageNet-LT with ResNet-50 · ImageNet-LT | Top-1 accuracy: 59.5 -> 59.7 | Helped |
| Use SHIKE rather than NCL (3N) | iNaturalist 2018 with ResNet-50 · iNaturalist 2018 | Top-1 accuracy: 74.9 -> 75.4 | Helped |
| Use SHIKE rather than NCL (3N) | Places-LT with ResNet-152 · Places-LT | accuracy (All): 41.8 -> 41.9 | Helped |
| Use SHIKE rather than PaCo | Places-LT · Places-LT few-shot | accuracy: 33.9 -> 44.8 | Helped |
| Add three experts (ablation) | Plain ResNet-32 · CIFAR100-LT IF100 | accuracy: 50.04 -> 54.23 | Helped |
| Add DKF to MoE (ablation) | Three-expert ResNet-32 · CIFAR100-LT IF100 | accuracy: 54.23 -> 55.26 | Helped |
| Add mutual knowledge distillation loss L_mu to MoE + DKF (ablation) | Three-expert ResNet-32 · CIFAR100-LT IF100 | accuracy: 55.26 -> 55.59 | Helped |
| Add non-target distillation loss L_nt (DKT) to MoE + DKF (ablation) | Three-expert ResNet-32 · CIFAR100-LT IF100 | accuracy: 55.26 -> 55.79 | Helped |
| Add L_mu and L_nt without adding DKF (ablation) | Three-expert ResNet-32 · CIFAR100-LT IF100 | accuracy: 54.23 -> 54.82 | Helped |
Closest papers by meaning
- Feature Fusion from Head to Tail for Long-Tailed Visual Recognition (AAAI 2024 · similarity 0.88)
- DiffuLT: Diffusion for Long-tail Recognition Without External Knowledge (NeurIPS 2024 · similarity 0.86)
- Decoupled Optimisation for Long-Tailed Visual Recognition (AAAI 2024 · similarity 0.86)
- From Sparse to Soft Mixtures of Experts (ICLR 2024 · similarity 0.85)
- Learning from Neighbors: Category Extrapolation for Long-Tail Learning (CVPR 2025 · similarity 0.85)
- Towards flexible perception with visual memory (ICML 2025 · similarity 0.85)
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.