Enhancing Object Coherence in Layout-to-Image Synthesis
Yibin Wang, Changhai Zhou, Honghui Xu
TL;DR
This work tackles object coherence in layout-to-image synthesis by introducing EOCNet, a diffusion-based model that jointly addresses semantic and physical coherence. It deploys Global Semantic Fusion (GSF) to fuse layout constraints with caption-based semantic relations, enabling caption-guided semantic control within the spatial layout, and Self-similarity Feature Enhancement (SFE) to integrate local physical coherence through a Rectified Cross Attention (RCA) and a Self-similarity Coherence Attention (SCA) that are fused for final generation. Empirical results on COCO-Stuff and ADE20K show state-of-the-art improvements in FID and Diversity Score, with ablations validating the effectiveness of both GSF and SFE in enhancing semantic control and physical texture coherence. The approach advances practical LIS by delivering higher fidelity images with better object-level controllability, leveraging caption inputs and local coherence cues to produce more coherent and realistic scenes.
Abstract
Layout-to-image synthesis is an emerging technique in conditional image generation. It aims to generate complex scenes, where users require fine control over the layout of the objects in a scene. However, it remains challenging to control the object coherence, including semantic coherence (e.g., the cat looks at the flowers or not) and physical coherence (e.g., the hand and the racket should not be misaligned). In this paper, we propose a novel diffusion model with effective global semantic fusion (GSF) and self-similarity feature enhancement modules to guide the object coherence for this task. For semantic coherence, we argue that the image caption contains rich information for defining the semantic relationship within the objects in the images. Instead of simply employing cross-attention between captions and latent images, which addresses the highly relevant layout restriction and semantic coherence requirement separately and thus leads to unsatisfying results shown in our experiments, we develop GSF to fuse the supervision from the layout restriction and semantic coherence requirement and exploit it to guide the image synthesis process. Moreover, to improve the physical coherence, we develop a Self-similarity Coherence Attention (SCA) module to explicitly integrate local contextual physical coherence relation into each pixel's generation process. Specifically, we adopt a self-similarity map to encode the physical coherence restrictions and employ it to extract coherent features from text embedding. Through visualization of our self-similarity map, we explore the essence of SCA, revealing that its effectiveness is not only in capturing reliable physical coherence patterns but also in enhancing complex texture generation. Extensive experiments demonstrate the superiority of our proposed method.