Rethinking Vector Field Learning for Generative Segmentation
This work addresses a key bottleneck in generative segmentation for computer vision researchers, offering incremental but impactful enhancements to existing diffusion-based methods.
The paper tackled the problem of slow convergence and poor class separation in diffusion models for generative segmentation by proposing a vector field reshaping strategy and a quasi-random category encoding scheme, resulting in significant improvements over vanilla flow matching approaches and narrowing the performance gap with discriminative methods.
Taming diffusion models for generative segmentation has attracted increasing attention. While existing approaches primarily focus on architectural tweaks or training heuristics, there remains a limited understanding of the intrinsic mismatch between continuous flow matching objectives and discrete perception tasks. In this work, we revisit diffusion segmentation from the perspective of vector field learning. We identify two key limitations of the commonly used flow matching objective: gradient vanishing and trajectory traversing, which result in slow convergence and poor class separation. To tackle these issues, we propose a principled vector field reshaping strategy that augments the learned velocity field with a detached distance-aware correction term. This correction introduces both attractive and repulsive interactions, enhancing gradient magnitudes near centroids while preserving the original diffusion training framework. Furthermore, we design a computationally efficient, quasi-random category encoding scheme inspired by Kronecker sequences, which integrates seamlessly with an end-to-end pixel neural field framework for pixel-level semantic alignment. Extensive experiments consistently demonstrate significant improvements over vanilla flow matching approaches, substantially narrowing the performance gap between generative segmentation and strong discriminative specialists.