Research map / Vision-language and multimodal / Multimodal LLMs
CountEx: Fine-Grained Counting via Exemplars and Exclusion
Yifeng Huang, Gia Khanh Nguyen, Minh Hoai Nguyen
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 Discriminative Query Refinement with exclusion prompts to LLMDet | LLMDet · CoCount test (novel-category NC) | MAE: 33.22 -> 26.61 | Improved |
| add Discriminative Query Refinement with exclusion prompts to LLMDet | LLMDet · CoCount test (known-category KC) | MAE: 16.82 -> 12.72 | Improved |
| subtract per-category density maps for negative exclusion instead of DQR | LLMDet · CoCount test (novel-category NC) | MAE: 33.22 -> 47.71 | Hurt |
| detection suppression for negatives instead of DQR | LLMDet · CoCount test (known-category KC) | MAE: 16.82 -> 16.94 | No clear effect |
| add selective query refinement stage (ablation) | CountEx · CoCount test (known-category KC) | MAE: 16.10 -> 12.72 | Improved |
| add shared feature identification stage (ablation) | CountEx · CoCount test (known-category KC) | MAE: 14.18 -> 12.72 | Improved |
| add exclusive feature extraction stage (ablation) | CountEx · CoCount test (known-category KC) | MAE: 13.40 -> 12.72 | Improved |
| add density prediction loss (ablation) | CountEx · CoCount test (known-category KC) | MAE: 15.20 -> 13.00 | Improved |
| add prototype losses (ablation) | CountEx · CoCount test (known-category KC) | MAE: 13.00 -> 12.72 | Improved |
| add negative text prompt (ablation) | CountEx · CoCount test (known-category KC) | MAE: 15.96 -> 13.22 | Improved |
| add negative exemplar boxes on top of negative text (ablation) | CountEx · CoCount test (known-category KC) | MAE: 13.22 -> 12.72 | Improved |
| use irrelevant negative prompt instead of none (ablation) | CountEx · CoCount test (novel-category NC) | MAE: 32.22 -> 28.43 | Improved |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 1 on most shared results.
- CountGD — CountGD: Multi-Modal Open-World Counting (NeurIPS 2024) · better on 3 of 4 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 0.87)
- Understanding Counting Mechanisms in Large Language and Vision-Language Models (CVPR 2026 · similarity 0.87)
- CountGD: Multi-Modal Open-World Counting (NeurIPS 2024 · similarity 0.87)
- OmniCount: Multi-label Object Counting with Semantic-Geometric Priors (AAAI 2025 · similarity 0.86)
- Unveiling the Visual Counting Bottleneck in Vision-Language Models (ICML 2026 · similarity 0.86)
- T2ICount: Enhancing Cross-modal Understanding for Zero-Shot Counting (CVPR 2025 · 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.
