Research map / Continual learning / Class-incremental
incDFM: Incremental Deep Feature Modeling for Continual Novelty Detection
Amanda Rios, Nilesh Ahuja, Ibrahima Ndiour, Utku Genc, Laurent Itti, Omesh Tickoo
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 incremental recruitment of confident novel samples (incDFM vs single-threshold DFM) (ablation) | DFM · CIFAR-10 class-incremental | AUROC (%): 74.0 -> 98.2 | Improved |
| add incremental recruitment of confident novel samples (ablation) | DFM · CIFAR-10 class-incremental | AUPR (%): 72.1 -> 98.2 | Improved |
| add incremental recruitment of confident novel samples (ablation) | DFM · EMNIST class-incremental | AUROC (%): 75.4 -> 98.7 | Improved |
| add incremental recruitment of confident novel samples (ablation) | DFM · offline CIFAR-10 -> SVHN | AUROC (%): 93.4 -> 99.9 | Improved |
| run two recruitment iterations instead of single thresholding (ablation) | DFM · CIFAR-10 | AUROC: 22% improvement | Improved |
| allow up to 10 maximum recruitment iterations (ablation) | DFM · CIFAR-10 | AUROC: up to 39.4% improvement | Improved |
| lower recruitment percentage per iteration (15%) (ablation) | incDFM · CIFAR-10 | iterations to converge | Hurt |
| use estimated novel samples instead of ground-truth novel samples for consolidation (ablation) | incDFM · CIFAR-10 | AUROC (%): 98.2 -> 98.2 | No clear effect |
| use estimated novel samples instead of ground-truth novel samples for consolidation (ablation) | ODIN · CIFAR-10 | AUROC (%): 87.3 -> 81.6 | Hurt |
| remove the task oracle | Softmax · 8-dataset continual novelty detection | AUROC (%): 99.9 -> 69.4 | Hurt |
| remove the task oracle | ODIN · 8-dataset continual novelty detection | AUROC (%): 99.8 -> 64.2 | Hurt |
| increase old:new imbalance from 1:1 to 1:4 | incDFM · CIFAR-10 | AUPR (%): 98.2 -> 95.0 | Hurt |
Closest papers by meaning
- Unsupervised Continual Anomaly Detection with Contrastively-Learned Prompt (AAAI 2024 · similarity 0.86)
- Demystifying Catastrophic Forgetting in Two-Stage Incremental Object Detector (ICML 2025 · similarity 0.85)
- One-for-More: Continual Diffusion Model for Anomaly Detection (CVPR 2025 · similarity 0.85)
- LoRanPAC: Low-rank Random Features and Pre-trained Models for Bridging Theory and Practice in Continual Learning (ICLR 2025 · similarity 0.84)
- H2ST: Hierarchical Two-Sample Tests for Continual Out-of-Distribution Detection (CVPR 2025 · similarity 0.84)
- AnaCP: Toward Upper-Bound Continual Learning via Analytic Contrastive Projection (NeurIPS 2025 · similarity 0.84)
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.
