Huan Wang

Junhan Zhu, Hesong Wang, Mingluo Su et al.

Large-scale text-to-image diffusion models, while powerful, suffer from prohibitive computational cost. Existing one-shot network pruning methods can hardly be directly applied to them due to the iterative denoising nature of diffusion models. To bridge the gap, this paper presents OBS-Diff, a novel one-shot pruning framework that enables accurate and training-free compression of large-scale text-to-image diffusion models. Specifically, (i) OBS-Diff revitalizes the classic Optimal Brain Surgeon (OBS), adapting it to the complex architectures of modern diffusion models and supporting diverse pruning granularity, including unstructured, N:M semi-structured, and structured (MHA heads and FFN neurons) sparsity; (ii) To align the pruning criteria with the iterative dynamics of the diffusion process, by examining the problem from an error-accumulation perspective, we propose a novel timestep-aware Hessian construction that incorporates a logarithmic-decrease weighting scheme, assigning greater importance to earlier timesteps to mitigate potential error accumulation; (iii) Furthermore, a computationally efficient group-wise sequential pruning strategy is proposed to amortize the expensive calibration process. Extensive experiments show that OBS-Diff achieves state-of-the-art one-shot pruning for diffusion models, delivering inference acceleration with minimal degradation in visual quality.

Jing Lou, Huan Wang, Longtao Chen et al.

In this paper, we will investigate the contribution of color names for the task of salient object detection. An input image is first converted to color name space, which is consisted of 11 probabilistic channels. By exploiting a surroundedness cue, we obtain a saliency map through a linear combination of a set of sequential attention maps. To overcome the limitation of only using the surroundedness cue, two global cues with respect to color names are invoked to guide the computation of a weighted saliency map. Finally, we integrate the above two saliency maps into a unified framework to generate the final result. In addition, an improved post-processing procedure is introduced to effectively suppress image backgrounds while uniformly highlight salient objects. Experimental results show that the proposed model produces more accurate saliency maps and performs well against twenty-one saliency models in terms of three evaluation metrics on three public data sets.