Research map / Detection and segmentation / Open-set and few-shot detection
Contextual Diversity for Active Learning
Sharat Agarwal, Himanshu Arora, Saket Anand, Chetan Arora
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 |
|---|---|---|---|
| use the entropy-weighted mixture instead of other mixture weights (ablation) | CDAL-RL · Cityscapes (25% data) | mIoU: 55.2 -> 56.3 | Improved |
| set reward weight alpha to 0.75 rather than 0.25 (ablation) | CDAL-RL · Cityscapes (25% data) | mIoU: 55.5 -> 56.3 | Improved |
| set reward weight alpha to 0.75 rather than 0.50 (ablation) | CDAL-RL · Cityscapes (25% data) | mIoU: 55.8 -> 56.3 | Improved |
| add visual and semantic representation rewards to the contextual diversity reward (ablation) | CDAL-RL · Cityscapes | mIoU | Improved |
| CD reward alone versus VAAL (ablation) | CDAL-RL · Cityscapes | mIoU | Improved |
| train the policy only once on a random 10% instead of every AL iteration (ablation) | CDAL-RL · Cityscapes | mIoU | Hurt |
| include CD reward for the Person class (ablation) | CDAL-RL · Cityscapes | per-class IoU | Improved |
| include CD reward for Person and Vegetation classes (ablation) | CDAL-RL · Cityscapes | per-class IoU | Improved |
| increase LSTM cell size from 256 to 1024 (ablation) | CDAL-RL · CIFAR-100 | accuracy | Improved |
| increase LSTM cell size from 256 to 1024 (ablation) | CDAL-RL · CIFAR-10 | accuracy | No clear effect |
| change budget step size from 5% to 10% (ablation) | CDAL-RL · CIFAR-100 | accuracy | No clear effect |
| use CDAL-RL frame selection instead of CEREALS full-frame mode | FCN8s on Cityscapes · Cityscapes (~10% data) | mIoU: 48.2 -> 51.6 | Improved |
Closest papers by meaning
- Improving Geo-diversity of Generated Images with Contextualized Vendi Score Guidance (ECCV 2024 · similarity 0.86)
- ReDAL: Region-Based and Diversity-Aware Active Learning for Point Cloud Semantic Segmentation (ICCV 2021 · similarity 0.85)
- Boundary Matters: A Bi-Level Active Finetuning Method (NeurIPS 2024 · similarity 0.85)
- Uncertainty Meets Diversity: A Comprehensive Active Learning Framework for Indoor 3D Object Detection (CVPR 2025 · similarity 0.84)
- Contextual Active Model Selection (NeurIPS 2024 · similarity 0.84)
- A Cross-Domain Benchmark for Active Learning (NeurIPS 2024 · similarity 0.83)
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.
