Research map / Detection and segmentation / Open-set and few-shot detection
People, Penguins and Petri Dishes: Adapting Object Counting Models to New Visual Domains and Object Types Without Forgetting
Read the paperCheck your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (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 shared multi-domain counter with refinement rather than the strongest listed baseline | Object counting · Shanghaitech Part A and Part B | MAE: 90.4 -> 85.7; 21.6 -> 17.7 | Helped |
| Use the shared multi-domain counter with refinement rather than the strongest listed baseline | Object counting · Penguins | MAE: 8.11 -> 5.8 | Helped |
| Use the shared multi-domain counter rather than Zhang et al. | Vehicle counting · TRANCOS | MAE: 4.2 -> 9.7 | Hurt |
| Use 100 × 100 rather than 50 × 50 patches (ablation) | Crowd counting regressor · Shanghaitech Part A validation | MAE: 100.8 -> 97.5 | Helped |
| Use 200 × 200 rather than 100 × 100 patches (ablation) | Crowd counting regressor · Shanghaitech Part A validation | MAE: 97.5 -> 158.5 | Hurt |
| Use MobileNet rather than VGG16 as feature extractor (ablation) | Crowd counting regressor · Shanghaitech Part A validation | MAE: 101.2 -> 94.1 | Helped |
| Prime on cells rather than crowds (ablation) | Sequential multi-domain counter · Shanghaitech Part A validation | MAE: 94.1 -> 90.2 | Helped |
| Adapt with domain-specific adapters rather than retraining the final 4 layers (ablation) | Cell-counting model adapted for crowd counting · Shanghaitech Part A validation | MAE: 128.3 -> 90.2 | Helped |
| Add fully convolutional prediction refinement (ablation) | Multi-domain counter · Crowds | MAE: 90.2 -> 86.5 | Helped |
| Add fully convolutional prediction refinement (ablation) | Multi-domain counter · Vehicles | MAE: 10.2 -> 10.1 | No clear effect |
| Add fully convolutional prediction refinement (ablation) | Multi-domain counter · Wildlife | MAE: 5.7 -> 5.6 | No clear effect |
| Add fully convolutional prediction refinement (ablation) | Multi-domain counter · Cells | MAE: 9.5 -> 8.4 | Helped |
Closest papers by meaning
- Learning to Count without Annotations (CVPR 2024 · similarity 0.86)
- Enhancing Low-Rank Adaptation with Recoverability-Based Reinforcement Pruning for Object Counting (AAAI 2025 · similarity 0.85)
- Vision Transformer Off-the-Shelf: A Surprising Baseline for Few-Shot Class-Agnostic Counting (AAAI 2024 · similarity 0.85)
- VLCounter: Text-Aware Visual Representation for Zero-Shot Object Counting (AAAI 2024 · similarity 0.85)
- T2ICount: Enhancing Cross-modal Understanding for Zero-Shot Counting (CVPR 2025 · similarity 0.84)
- OmniCount: Multi-label Object Counting with Semantic-Geometric Priors (AAAI 2025 · 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.