CoLoGen: Progressive Learning of Concept-Localization Duality for Unified Image Generation
This work addresses the challenge of unified image generation for AI and computer vision applications, offering a principled representational perspective, though it appears incremental as it builds on existing diffusion frameworks.
The paper tackles the problem of unified conditional image generation by addressing the representational conflict between conceptual understanding and localization cues, proposing CoLoGen, a diffusion framework that progressively learns and reconciles this duality, achieving competitive or superior performance in editing, controllable generation, and customized generation tasks.
Unified conditional image generation remains difficult because different tasks depend on fundamentally different internal representations. Some require conceptual understanding for semantic synthesis, while others rely on localization cues for spatial precision. Forcing these heterogeneous tasks to share a single representation leads to concept-localization representational conflict. To address this issue, we propose CoLoGen, a unified diffusion framework that progressively learns and reconciles this concept-localization duality. CoLoGen uses a staged curriculum that first builds core conceptual and localization abilities, then adapts them to diverse visual conditions, and finally refines their synergy for complex instruction-driven tasks. Central to this process is the Progressive Representation Weaving (PRW) module, which dynamically routes features to specialized experts and stably integrates their outputs across stages. Experiments on editing, controllable generation, and customized generation show that CoLoGen achieves competitive or superior performance, offering a principled representational perspective for unified image generation.