Research map / Label-efficient and robust learning / Recognition and long-tail
Learning to Discover Novel Visual Categories via Deep Transfer Clustering
Kai Han, Andrea Vedaldi, Andrew Zisserman
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 consistency constraint (Pi-model) to DEC-based DTC (ablation) | DTC-Baseline · CIFAR-10 | ACC: 74.9% -> 87.5% | Improved |
| add consistency constraint (Pi-model) to DEC-based DTC (ablation) | DTC-Baseline · SVHN | ACC: 57.6% -> 60.9% | Improved |
| add Pi consistency constraint to DTC (ablation) | DTC-Baseline · CIFAR-100 | ACC: 72.1% -> 70.6% | Hurt |
| add temporal ensembling (TE) to DTC (ablation) | DTC-Baseline · CIFAR-100 | ACC: 72.1% -> 72.8% | Improved |
| add temporal ensembling (TE) to DTC (ablation) | DTC-Baseline · SVHN | ACC: 57.6% -> 55.8% | Hurt |
| add temporal ensemble targets (TEP) to DTC (ablation) | DTC-Baseline · CIFAR-10 | ACC: 74.9% -> 75.2% | Mixed |
| add Pi consistency constraint to DTC (ablation) | DTC-Baseline · OmniGlot | ACC: 87.9% -> 89.0% | Improved |
| use estimated number of categories for KCL (ablation) | KCL · ImageNet (unknown K) | ACC: 65.2% -> 71.4% | Improved |
| use estimated number of categories for MCL (ablation) | MCL · OmniGlot (unknown K) | ACC: 80.2% -> 80.5% | Improved |
| use estimated number of categories for MCL | MCL · ImageNet (unknown K) | ACC: 71.5% -> 72.9% | Improved |
| transfer from labelled-class feature extractor, DTC vs KCL | KCL · CIFAR-10 | ACC: 66.5% -> 87.5% | Improved |
| estimate unknown category number instead of assuming K | DTC · OmniGlot | ACC: 89.0% -> 87.0% | Hurt |
Closest papers by meaning
- Automatically Discovering and Learning New Visual Categories with Ranking Statistics (ICLR 2020 · similarity 0.90)
- AutoNovel: Automatically Discovering and Learning Novel Visual Categories (TPAMI 2021 · similarity 0.90)
- Automatically Discovering Novel Visual Categories With Adaptive Prototype Learning (TPAMI 2023 · similarity 0.89)
- Semantic-Guided Novel Category Discovery (AAAI 2024 · similarity 0.88)
- Debiased Novel Category Discovering and Localization (AAAI 2024 · similarity 0.87)
- Deep Taxonomic Networks for Unsupervised Hierarchical Prototype Discovery (NeurIPS 2025 · similarity 0.87)
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.
