COLLAR: Cascaded Object-Level Latent Refinement for High-Fidelity Conditional Generation
For researchers working on conditional image generation, COLLAR addresses visual artifacts and imprecise control in object-level generation without requiring additional training.
COLLAR is a training-free framework that improves object-level control in Diffusion Transformers by progressively refining features via Field-of-View expansion, achieving state-of-the-art performance on COCO-MIG and COCO-POS benchmarks in semantic alignment, image quality, and spatial fidelity.
Achieving high-fidelity object-level control in Diffusion Transformers remains a significant challenge despite the introduction of structural priors like depth and Canny maps. Current object-level conditional generation methods frequently suffer from visual artifacts and struggle to maintain precise control over objects within small localized regions. To address these limitations, we propose Cascaded Object-Level Latent Refinement (COLLAR), a training-free framework that progressively optimizes object-level features via the Field-of-View (FoV) expansion. First, we propose the Cross-Scale Semantic Alignment (CSSA) module to address spatial-semantic gaps by injecting object-level features into extended-FoV branches via attention mechanisms. To further optimize these features, the Cyclic Feature Injection (CFI) module introduces a reciprocal background feedback mechanism. It leverages a frequency-based adaptive strategy to selectively update the global backbone with context-aligned local information. Finally, the extended-FoV branch serves as a hub for feature optimization, ensuring that object-level features are integrated into the global generation process without compromising final image quality. Extensive experiments on the COCO-MIG and COCO-POS benchmarks demonstrate that our approach consistently outperforms state-of-the-art methods across semantic alignment, image quality, and spatial fidelity.