Class Agnostic Counting Dataset

Class-agnostic counting aims to automatically count objects of any class without prior training on that specific class, a significant challenge with broad applications in diverse fields. Recent research focuses on developing training-free or zero-shot methods, often employing prompt-based approaches that leverage various input modalities (e.g., bounding boxes, points, text) and incorporate context-aware mechanisms to improve accuracy, particularly in complex scenes. These advancements address limitations of existing datasets and models, leading to improved performance and generalizability across different object types and datasets, thereby reducing the reliance on extensive human annotation.