YOLO-Count: Differentiable Object Counting for Text-to-Image Generation

TL;DR

YOLO-Count introduces a differentiable object counting model that improves text-to-image generation by providing accurate counts and control over object quantities through a novel regression target and hybrid supervision.

Contribution

It presents YOLO-Count, a novel differentiable counting model with a 'cardinality' map for open-vocabulary counting and precise control in T2I generation.

Findings

Achieves state-of-the-art counting accuracy.

Enables robust quantity control in T2I systems.

Demonstrates effective gradient-based optimization.

Abstract

We propose YOLO-Count, a differentiable open-vocabulary object counting model that tackles both general counting challenges and enables precise quantity control for text-to-image (T2I) generation. A core contribution is the 'cardinality' map, a novel regression target that accounts for variations in object size and spatial distribution. Leveraging representation alignment and a hybrid strong-weak supervision scheme, YOLO-Count bridges the gap between open-vocabulary counting and T2I generation control. Its fully differentiable architecture facilitates gradient-based optimization, enabling accurate object count estimation and fine-grained guidance for generative models. Extensive experiments demonstrate that YOLO-Count achieves state-of-the-art counting accuracy while providing robust and effective quantity control for T2I systems.