FlashClear: Ultra-Fast Image Content Removal via Efficient Step Distillation and Feature Caching

TL;DR

FlashClear is a highly efficient diffusion-based image content removal method that significantly accelerates inference by focusing on foreground regions, achieving up to 122x speedup without sacrificing quality.

Contribution

The paper introduces RAD and FPAC strategies for rapid, region-aware diffusion-based object removal, reducing computational cost and inference time.

Findings

FlashClear achieves up to 8.26× speedup over ObjectClear.

FlashClear maintains high visual quality and fidelity.

The framework outperforms existing methods on the OBER benchmark.

Abstract

Recently, diffusion-based object removal models have achieved impressive results in eliminating objects and their associated visual effects. However, they indiscriminately denoise all tokens across all timesteps, ignoring that removal usually involves small foreground regions. This strategy introduces substantial computational overhead and prolonged inference times. To overcome this computational burden, we propose a latent discriminator to implement Region-aware Adversarial Distillation (RAD), yielding a highly efficient few-step model named FlashClear. Furthermore, tailored to few-step diffusion models, we propose FPAC (Foreground-Prioritized Asymmetric Attention and Caching), a training-free acceleration strategy. Extensive experiments demonstrate that our framework provides massive acceleration while maintaining or exceeding the performance of our base model, ObjectClear. Notably, on the OBER benchmark, our FlashClear achieves up to 8.26$\times$ and 122$\times$ speedup over ObjectClear and OmniPaint, respectively, while maintaining high visual quality and fidelity.