ReasonX: MLLM-Guided Intrinsic Image Decomposition
This work addresses the challenge of applying intrinsic image decomposition to diverse, real-world images, which is important for computer vision applications, though it appears incremental as it builds on existing methods with a novel supervision approach.
The paper tackled the problem of generalizing intrinsic image decomposition to real-world scenarios by proposing ReasonX, a framework that uses a multimodal large language model as a perceptual judge to provide comparative supervision, resulting in improvements such as 9-25% WHDR reduction on IIW albedo and up to 46% depth accuracy gains on ETH3D.
Intrinsic image decomposition aims to separate images into physical components such as albedo, depth, normals, and illumination. While recent diffusion- and transformer-based models benefit from paired supervision from synthetic datasets, their generalization to diverse, real-world scenarios remains challenging. We propose ReasonX, a novel framework that leverages a multimodal large language model (MLLM) as a perceptual judge providing relative intrinsic comparisons, and uses these comparisons as GRPO rewards for fine-tuning intrinsic decomposition models on unlabeled, in-the-wild images. Unlike RL methods for generative models, our framework aligns conditional intrinsic predictors by rewarding agreement between the judge's relational assessments and analytically derived relations from the model's outputs. ReasonX is model-agnostic and can be applied to different intrinsic predictors. Across multiple base architectures and modalities, ReasonX yields significant improvements, including 9-25% WHDR reduction on IIW albedo and up to 46% depth accuracy gains on ETH3D, highlighting the promise of MLLM-guided comparative supervision to bridge low- and high-level vision reasoning.