Research map / Vision-language and multimodal / CLIP and zero-shot
Enhancing Zero-shot Object Counting via Text-guided Local Ranking and Number-evoked Global Attention
Shiwei Zhang, Qi Zhou, Wei Ke
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 Text-guided Local Ranking (ablation) | ClipCount · FSC-147 test | MAE: 17.45 -> 16.71 | Improved |
| add Number-evoked Global Attention (ablation) | ClipCount · FSC-147 test | MAE: 17.45 -> 16.81 | Improved |
| add both TLR and NGA (ablation) | ClipCount · FSC-147 val | MAE: 18.93 -> 17.52 | Improved |
| use coarse-to-fine number prompts instead of coarse only (ablation) | ClipCount+Ours · FSC-147 val | MAE: 18.17 -> 17.52 | Improved |
| apply strategy to ClipCount | ClipCount · FSC-147 test | RMSE: 104.88 -> 100.63 | Improved |
| apply strategy to VLCounter | VLCounter · FSC-147 test | RMSE: 106.16 -> 107.47 | Hurt |
| apply strategy to GDino (w/ finetune) | GDino finetune · FSC-147 test | RMSE: 104.98 -> 106.5 | Hurt |
| apply strategy to CountGD | CountGD · FSC-147 test | MAE: 14.76 -> 12.84 | Improved |
| apply strategy to GDino finetune | GDino finetune · ZSC-8K val | MAE: 13.92 -> 11.17 | Improved |
| apply strategy to ClipCount | ClipCount · ZSC-8K testlongtail | MAE: 22.65 -> 21.90 | Improved |
| apply strategy to ClipCount, evaluate cross-dataset | ClipCount · ShanghaiTech Part A (from FSC-147) | MAE: 192.6 -> 176.24 | Improved |
| set delta to 0.5 instead of 0.1 or 0.9 (ablation) | ClipCount+Ours · FSC-147 | MAE | Improved |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 2 of 2 on most shared results.
- VLCounter — VLCounter: Text-Aware Visual Representation for Zero-Shot Object Counting (AAAI 2024) · better on 2 of 2 shared results
- CountGD — CountGD: Multi-Modal Open-World Counting (NeurIPS 2024) · better on 2 of 2 shared results
Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.
Closest papers by meaning
- VLCounter: Text-Aware Visual Representation for Zero-Shot Object Counting (AAAI 2024 · similarity 0.92)
- T2ICount: Enhancing Cross-modal Understanding for Zero-Shot Counting (CVPR 2025 · similarity 0.91)
- CountGD: Multi-Modal Open-World Counting (NeurIPS 2024 · similarity 0.89)
- Point Segment and Count: A Generalized Framework for Object Counting (CVPR 2024 · similarity 0.88)
- Single Domain Generalization for Few-Shot Counting via Universal Representation Matching (CVPR 2025 · similarity 0.88)
- Vision Transformer Off-the-Shelf: A Surprising Baseline for Few-Shot Class-Agnostic Counting (AAAI 2024 · similarity 0.87)
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.
