Research map / Label-efficient and robust learning / Recognition and long-tail
Beyond the Static World: Continual Category Discovery under Visual Drift
Wei Feng, Yiwen Jiang, Sijin Zhou, Zongyuan Ge
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 WSM and CDSA to the baseline (ablation) | OCCD framework · DomainNet Real->Painting, Painting domain (All) | clustering accuracy (%): 28.7 -> 34.1 | Improved |
| add WSM and CDSA to the baseline (ablation) | OCCD framework · DomainNet Real->Painting, Real domain (All) | clustering accuracy (%): 54.6 -> 57.7 | Improved |
| add CTCC to WSM + CDSA (ablation) | OCCD framework · DomainNet Real->Painting, Painting domain (All) | clustering accuracy (%): 34.1 -> 38.9 | Improved |
| add CTCC to WSM + CDSA (ablation) | OCCD framework · DomainNet Real->Painting, Painting domain (New) | clustering accuracy (%): 31.9 -> 37.7 | Improved |
| add CDSA to WSM + CTCC (ablation) | OCCD framework · DomainNet Real->Painting, Painting domain (All) | clustering accuracy (%): 32.8 -> 38.9 | Improved |
| use PUOT-based weight-aware separation instead of entropy-based (ablation) | OCCD framework · DomainNet Real->Painting, Painting domain (All) | clustering accuracy (%): 29.4 -> 38.9 | Improved |
| use PUOT-based weight-aware separation instead of energy-based (ablation) | OCCD framework · DomainNet Real->Painting, Painting domain (All) | clustering accuracy (%): 30.1 -> 38.9 | Improved |
| set both loss weights lambda1 and lambda2 to zero (ablation) | OCCD framework · DomainNet Real->Painting | clustering accuracy | Hurt |
| vary lambda1 and lambda2 within a reasonable non-zero range (ablation) | OCCD framework · DomainNet Real->Painting | clustering accuracy | No clear effect |
| use HiLo patch-mixing augmentation for distribution shift in GCD | GCD methods · CUB-C, Original domain (All) | clustering accuracy (%): 29.4 -> 32.3 | Improved |
Closest papers by meaning
- Continual Generalized Category Discovery: Learning and Forgetting from a Bayesian Perspective (ICML 2025 · similarity 0.92)
- Happy: A Debiased Learning Framework for Continual Generalized Category Discovery (NeurIPS 2024 · similarity 0.91)
- Consistent Supervised-Unsupervised Alignment for Generalized Category Discovery (NeurIPS 2025 · similarity 0.89)
- Generalized Category Discovery under Domain Shift: A Frequency Domain Perspective (NeurIPS 2025 · similarity 0.89)
- Prior-Constrained Association Learning for Fine-Grained Generalized Category Discovery (AAAI 2025 · similarity 0.89)
- Transfer and Alignment Network for Generalized Category Discovery (AAAI 2024 · 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.
