Search papers, labs, and topics across Lattice.
The paper introduces CoLoGen, a unified diffusion framework designed to address the concept-localization representational conflict in conditional image generation by progressively learning and reconciling these dual aspects. CoLoGen employs a staged curriculum that builds conceptual and localization abilities, adapts them to diverse conditions, and refines their synergy for complex tasks. The core of CoLoGen is the Progressive Representation Weaving (PRW) module, which dynamically routes features to specialized experts and integrates their outputs across stages.
Achieve unified image generation by progressively disentangling and weaving together concept and localization representations within a diffusion framework, outperforming prior methods on diverse 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.
