OverLay++: Dense-Overlap Layout-to-Image Generation Dataset

2026-10-06

AI Summary

This study addresses the bottleneck of insufficient high-density, multi-object interaction training data for layout-to-image generation in complex overlapping scenes. To this end, we construct a large-scale dataset comprising 500,000 images with an average of 6.6 objects per image, and propose a streamlined automated data generation pipeline that integrates dense detection with fine-grained semantic annotations to achieve high-quality synthesis. Compared to existing methods, our dataset increases annotation density by 1.67 times and description length sixfold, effectively bridging the data gap for complex scenes. Experimental results demonstrate that state-of-the-art models trained on this dataset exhibit significantly improved performance and accelerated convergence, validating the critical role of dense overlapping supervision in controllable image generation.

Abstract

Layout-to-Image generation has made substantial progress in spatial and object-level control. However, existing methods still struggle with complex scenes containing many overlapping and interacting objects. We argue that training data is a particular bottleneck: existing datasets lack examples with dense, complex object interactions. To address this gap, we introduce OverLay++, a large-scale Layout-to-Image dataset with structurally complex scenes. OverLay++ contains approximately 500K images with an average of 6.6 objects per image, exceeding existing datasets by 1.67 times in annotation density. Beyond annotation density, OverLay++ provides rich semantic detail with object captions over six times longer than in current datasets. Our dataset generation pipeline is simple and produces dense, overlapping object annotations with rich per-object captions. Across multiple benchmarks, state-of-the-art Layout-to-Image methods trained on the OverLay++ dataset show consistent improvement and faster convergence, demonstrating the importance of dense, overlap-aware, and caption-rich supervision for controllable image generation.

Problem

• Layout-to-Image Generation
• Dense Overlap
• Complex Scenes
• Training Data Bottleneck

Innovation

• Layout-to-Image Generation
• Dense-Overlap Dataset
• Data Generation Pipeline
• Controllable Image Generation
• Rich Object Captions