Research map / Label-efficient and robust learning / Semi-supervised learning
Realistic Evaluation of Deep Active Learning for Image Classification and Semantic Segmentation
Sudhanshu Mittal, Joshua Niemeijer, Özgün Çiçek, Maxim Tatarchenko, Jan Ehrhardt, Jörg P. Schäfer, Heinz Handels A, Thomas Brox
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 strong data augmentation (ablation) | Random sampling · CIFAR-10 | accuracy: 76.43 -> 89.55 | Improved |
| add strong augmentation (ablation) | AL methods · CIFAR-10 | accuracy: within a range of 0.4% | No clear effect |
| add LL active learning without augmentation (ablation) | Random baseline · CIFAR-10 | accuracy: 76.43 -> 78.71 | Improved |
| add semi-supervised learning (ablation) | Random sampling · CIFAR-10 | accuracy: 89.55 -> 94.11 | Improved |
| add semi-supervised learning (ablation) | Random sampling · CIFAR-100 | accuracy: 61.67 -> 68.14 | Improved |
| add LL query to SSL (ablation) | SSL-Random · CIFAR-100 | accuracy: 68.14 -> 66.99 | Hurt |
| use AL instead of random in low budget (250/500) (ablation) | SSL · CIFAR-10/100 | accuracy | No clear effect |
| add batch Coreset with SSL (ablation) | A2D2-3k · A2D2-3k | mIoU: 57.57 -> 61.13 | Improved |
| add entropy acquisition over random (ablation) | Random · A2D2-3k | mIoU: 56.14 -> 60.16 | Improved |
| add SSL to Entropy (ablation) | Entropy · Cityscapes | mIoU: 61.83 -> 60.61 | Hurt |
| add SSL to Coreset (ablation) | Coreset · Cityscapes | mIoU: 60.89 -> 63.14 | Improved |
| use Entropy instead of Random (ablation) | Random · BCSS 1-1 low budget | mIoU@3: 56.29 -> 53.67 | Hurt |
Closest papers by meaning
- Integrating Deep Metric Learning with Coreset for Active Learning in 3D Segmentation (NeurIPS 2024 · similarity 0.90)
- Navigating the Pitfalls of Active Learning Evaluation: A Systematic Framework for Meaningful Performance Assessment (NeurIPS 2023 · similarity 0.89)
- Reinforced active learning for image segmentation (ICLR 2020 · similarity 0.89)
- Diffusion-Driven Two-Stage Active Learning for Low-Budget Semantic Segmentation (NeurIPS 2025 · similarity 0.88)
- Active Label Correction for Semantic Segmentation with Foundation Models (ICML 2024 · similarity 0.87)
- A Cross-Domain Benchmark for Active Learning (NeurIPS 2024 · similarity 0.86)
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.
