Research map / Detection and segmentation / Open-set and few-shot detection
Interactive Class-Agnostic Object Counting
Yifeng Huang, Viresh Ranjan, Minh Hoai
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 five interaction iterations instead of five additional exemplars | FamNet · FSC-147 test | MAE: 21.52 -> 11.75 | Improved |
| add channel-wise refinement to spatial-wise refinement (ablation) | refinement module · FSC-147 val | MAE: 21.63 -> 12.79 | Improved |
| replace IPSE density segmentation with K-means, MSER, Watershed or DBSCAN (ablation) | FamNet · FSC-147 test | MAE: 11.75 -> 15.32 (K-means best alternative) | Hurt |
| add global loss to local loss (ablation) | adaptation loss · FSC-147 val | MAE: 15.41 -> 13.25 | Improved |
| add confidence scaling on FamNet (ablation) | adaptation loss · FSC-147 val | MAE: 13.25 -> 12.79 | Improved |
| adapt only refinement module instead of all regression head parameters (ablation) | FamNet · FSC-147 val | MAE: 17.07 -> 12.79 | Improved |
| select region with largest error instead of random (ablation) | user feedback simulation · FSC-147 val | MAE: 12.79 -> 10.61 | Improved |
| prioritize background regions instead of random (ablation) | user feedback simulation · FSC-147 val | MAE: 12.79 -> 14.02 | Hurt |
| give interactions consecutively rather than non-consecutively (ablation) | FamNet · FSC-147 val | MAE: 16.87 -> 12.79 | Improved |
| insert refinement module after first conv rather than deeper or on correlation map (ablation) | FamNet · FSC-147 val | MAE: 13.87 -> 12.79 | Improved |
| apply framework to DM-Count crowd counter with adjusted ranges | DM-Count · ShanghaiTech A | MAE: 59.60 -> 33.85 | Improved |
Closest papers by meaning
- A Fixed-Point Approach to Unified Prompt-Based Counting (AAAI 2024 · similarity 0.89)
- Learning to Count without Annotations (CVPR 2024 · similarity 0.87)
- Vision Transformer Off-the-Shelf: A Surprising Baseline for Few-Shot Class-Agnostic Counting (AAAI 2024 · similarity 0.86)
- DAVE - A Detect-and-Verify Paradigm for Low-Shot Counting (CVPR 2024 · similarity 0.86)
- Referring Expression Counting (CVPR 2024 · similarity 0.86)
- Point Segment and Count: A Generalized Framework for Object Counting (CVPR 2024 · similarity 0.85)
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.
