Research map / Label-efficient and robust learning / Semi-supervised learning
A Unified Objective for Novel Class Discovery
Enrico Fini, Enver Sangineto, Stéphane Lathuilière, Zhun Zhong, Moin Nabi, Elisa Ricci
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 logit concatenation into one shared softmax (unified objective) (ablation) | UNO · CIFAR10 task-agnostic all classes | accuracy: 65.9 -> 93.4 | Improved |
| add logit concatenation into one shared softmax (unified objective) (ablation) | UNO · CIFAR100-50 task-agnostic all classes | accuracy: 54.3 -> 61.1 | Improved |
| add overclustering heads (ablation) | UNO · CIFAR10 task-agnostic all classes | accuracy: 92.0 -> 93.4 | Improved |
| add overclustering heads (ablation) | UNO · CIFAR100-50 task-agnostic all classes | accuracy: 59.0 -> 61.1 | Improved |
| add overclustering heads (ablation) | UNO · CIFAR100-50 task-aware labeled classes | accuracy: 78.9 -> 78.8 | No clear effect |
| use strong instead of weak augmentation (ablation) | UNO · CIFAR10 task-agnostic all classes | accuracy: 91.8 -> 93.4 | Improved |
| use strong instead of weak augmentation (ablation) | UNO · CIFAR100-50 task-agnostic all classes | accuracy: 59.1 -> 61.1 | Improved |
| apply strong augmentations to RS | RS · novel class discovery | clustering accuracy | No clear effect |
| swapped prediction instead of averaging logits (ablation) | UNO · CIFAR10 (train) | clustering accuracy: 94.4 -> 96.1 | Improved |
| swapped prediction instead of averaging logits (ablation) | UNO · CIFAR100-50 (train) | clustering accuracy: 52.2 -> 52.9 | Improved |
| swapped prediction instead of averaging pseudo-labels (ablation) | UNO · CIFAR100-50 (train) | clustering accuracy: 52.5 -> 52.9 | No clear effect |
| self-supervised pretraining instead of supervised pretraining (ablation) | UNO · novel class discovery | clustering accuracy | No clear effect |
Closest papers by meaning
- Generalized Category Discovery (CVPR 2022 · similarity 0.90)
- Semantic-Guided Novel Category Discovery (AAAI 2024 · similarity 0.89)
- OpenMix: Reviving Known Knowledge for Discovering Novel Visual Categories in an Open World (CVPR 2021 · similarity 0.89)
- Automatically Discovering Novel Visual Categories With Adaptive Prototype Learning (TPAMI 2023 · similarity 0.89)
- Novel Class Discovery in Semantic Segmentation (CVPR 2022 · similarity 0.89)
- Debiased Novel Category Discovering and Localization (AAAI 2024 · 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.
