Collaborative Distillation for Ultra-Resolution Universal Style Transfer

Wang, Huan; Li, Yijun; Wang, Yuehai; Hu, Haoji; Yang, Ming–Hsuan

doi:10.1109/cvpr42600.2020.00193

Cited by 96 publications

(84 citation statements)

References 33 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…As shown in Table Ⅱ, the proposed algorithm achieves the best performance on shape retention. And its performance on style migration is only a little inferior to [14,16], but the performance of [14,16] on shape retention is the least satisfactory. As shown in Figure 4, it is difficult to figure out the contents in the stylized images generated by [14,16], while the proposed algorithm does achieve high performance both on shape retention and style migration.…”

Section: Comparision With Other Sota Algorithmsmentioning

confidence: 99%

“…And its performance on style migration is only a little inferior to [14,16], but the performance of [14,16] on shape retention is the least satisfactory. As shown in Figure 4, it is difficult to figure out the contents in the stylized images generated by [14,16], while the proposed algorithm does achieve high performance both on shape retention and style migration. Besides, as shown in Figure 2 (recommend to zoom in), the good performance on shape retention makes the proposed algorithm has the potential to be applied on photo realistic style transfer.…”

Section: Comparision With Other Sota Algorithmsmentioning

confidence: 99%

“…The soaring performance of style transfer is attributed to Gatys et al introducing neural network into style transfer [5], which outperforms the non-parametric algorithms [6,7]. Then, large numbers of researches on style transfer spring out [8]- [19], such as photo-realistic style transfer [8,9], migrating the style of an artist [10,11], improving the quality of style transfer [12]- [16], and real-time style transfer [17]- [19].…”

Section: Introductionmentioning

confidence: 99%

“…One of the main challenge in style transfer is how to fully migrate style while maintaining content consistent (shape retention). As shown in Figure 1, the recent state-of-the-art (SOTA) algorithms fail to achieve satisfactory performance on shape retention and suffer from content distortion [12][13][14][15][16].…”

Section: Introductionmentioning

confidence: 99%

“…However, the SOTA algorithms consist of complex networks [12]- [14] or need additional training when style changes [15,17,18]. Wang et al [16] apply knowledge distillation [19,20] to replace the VGG [21] in [14] with a small network, where computational cost is largely reduced, but content distortion still exists.…”

Section: Introductionmentioning

confidence: 99%

See 4 more Smart Citations

Fast Style Transfer with High Shape Retention

Huang

Ling

Jing

et al. 2021

2021 IEEE International Symposium on Circuits and Systems (ISCAS)

View full text Add to dashboard Cite

Since deep learning was introduced into style transfer, remarkable results have been achieved in it and it is widely used in multimedia fields, such as photography. However, the computational costs of the existing state-of-the-art (SOTA) arbitrary style transfer algorithms are still too complex to apply them on mobile device and high resolution, and their performance on shape retention is not satisfactory enough. To deal with the above problems, we propose a novel arbitrary style transfer algorithm. Specially, we propose a new network which ensures the low computational cost and high shape retention. Moreover, we propose the weighted style loss function to improve the performance on style migration. The experimental results show that the proposed algorithm achieves better results with lower computational cost than the SOTA algorithms.

show abstract

Section: Comparision With Other Sota Algorithmsmentioning

confidence: 99%

Section: Comparision With Other Sota Algorithmsmentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

See 3 more Smart Citations

Fast Style Transfer with High Shape Retention

Huang

Ling

Jing

et al. 2021

2021 IEEE International Symposium on Circuits and Systems (ISCAS)

View full text Add to dashboard Cite

show abstract

PKD‐Net: Distillation of prior knowledge for image completion by multi‐level semantic attention

Lu,

Lin,

Xing

et al. 2023

Concurrency and Computation

View full text Add to dashboard Cite

SummaryPrior knowledge plays a crucial role in image completion. Although almost all of the existing image completion methods use prior knowledge to complete the image to be repaired from different perspectives, the learning and modeling of the prior knowledge is still a challenging problem. In order to address this issue, we propose a novel prior knowledge distillation framework (PKD‐Net) which could distill prior knowledge of structure and style from multiple semantic space and generates not only plausible content but also consistent style with surrounding image area. Our PKD‐Net replaces the skip connection in the vanilla U‐Net with a semantic shift attention module. The semantic shift attention module takes features from encoder layer and those from decoder layer as input pairs to output shifted features which take into account the long‐range dependency of encoder layer features and corresponding decoder layer features from the perspective of local structure and style. Semantic shift attention module models the global interdependencies in local spatial structures (patches centered at each position) and style (appearance texture) dimensions respectively, which could implement distillation of prior knowledge from two aspects: structure and style. Experiments on multiple datasets including faces (CelebA, CelebA‐HQ) and natural images (ImageNet, Places2, Paris Street View) demonstrate that our proposed approach generates higher quality completion results than existing ones.

show abstract