🍒Count Anything, Any Granularity🍒 👉Open-world counting as multi-grained counting, where...

2026-05-12 · Source: AI with Papers - Artificial Intelligence & Deep Learning (@AI_DeepLearning) - Telegram · Field: Technology & Digital — Artificial Intelligence & Machine Learning, Computer Vision · Depth: Expert, quick

Summary

A new open-world counting method, "Count Anything, Any Granularity" (CAAG), enables multi-grained counting by using visual exemplars to define target appearance and fine-grained text to specify semantic granularity. This approach supports counting across five explicit levels of detail, allowing users to precisely control what is counted and at what resolution. The project includes a public repository and data, both released under an Apache license, facilitating broader research and application development in computer vision and object counting. This system addresses the challenge of flexible and adaptable counting in diverse visual environments.

Key takeaway

For research scientists developing computer vision applications, CAAG offers a robust framework for highly customizable object counting. You should explore its multi-grained capabilities to enhance precision in scenarios requiring specific semantic granularity, moving beyond traditional fixed-category counting. Consider integrating this Apache-licensed method into your projects for flexible object enumeration.

Key insights

CAAG enables flexible, multi-grained object counting using visual exemplars and fine-grained text specifications.

Principles

• Combine visual and textual cues for counting.
• Support explicit semantic granularity levels.

Method

The method uses visual exemplars to define target objects and text descriptions to specify the desired semantic granularity across five distinct levels for open-world counting.

In practice

• Count objects with varying levels of detail.
• Apply to diverse visual environments.

Topics

• Open-world Counting
• Multi-grained Counting
• Visual Exemplars
• Semantic Granularity
• Object Counting

Best for: Research Scientist, AI Scientist, Computer Vision Engineer, Machine Learning Engineer

Related on AIssential

• Does it Really Count? Assessing Semantic Grounding in Text-Guided Class-Agnostic Counting — Current class-agnostic counting models struggle with semantic grounding, leading to unreliable object class identification from text prompts.
• Every9D-21M: Large-Scale Real-World 9D Canonicalization of Everyday Objects — Large-scale, real-world 9D pose datasets can be built efficiently by propagating sparse manual annotations across object-centric videos.
• FruitEnsemble: MLLM-Guided Arbitration for Heterogeneous ensemble in Fine-Grained Fruit Recognition — A two-stage dynamic inference framework combines heterogeneous ensembles with MLLM-guided arbitration for fine-grained fruit recognition.
• Test-Time Training for Robust Text-Guided Open-Vocabulary Object Counting — Dual-TTT enhances text-guided open-vocabulary object counting robustness in degraded conditions by test-time training a lightweight denoising module.
• Multi-Modal Hyper-Graph Fusion for Low-Light Crowd Counting — Multi-modal hyper-graph fusion significantly improves low-light crowd counting by integrating RGB, depth, and edge cues.
• CAOA -- Completion-Assisted Object-CAD Alignment — CAOA integrates point cloud completion and symmetry-aware pose estimation to improve object-CAD alignment in noisy real-world scans.

Open in AIssential →

Editorial summary, takeaway, and curation by AIssential. Original article published by AI with Papers - Artificial Intelligence & Deep Learning (@AI_DeepLearning) - Telegram.