AI Paper Atlas

Research map / Image generation / Editing and style transfer

Decoder Network Over Lightweight Reconstructed Feature for Fast Semantic Style Transfer

Ming Lu, Hao Zhao, Anbang Yao, Feng Xu, Yurong Chen, Li Zhang

ICCV 2017 · Image generation › Editing and style transfer · approach: Arbitrary neural style transfer · 51 citations (OpenAlex)

Read the paper

AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.

Measured effects reported in the paper

What was changedWhereMeasuredEffect
replace backward optimization through the network with feature-space reconstruction plus decodersemantic style transfer · Style transfer without segmentation masks (50 samples)average time (s): 112s -> 1.35sImproved
use feed-forward transfer net instead of the proposed methodstyle transfer · Style transfer without segmentation masks (50 samples)average time (s): 1.35s -> 0.063sImproved
replace Gatys2016 with the proposed methodportrait style transfer · Portrait user studyaverage time per image (s): 127.3s -> 1.48sImproved
replace Gatys2016 with the proposed methodportrait style transfer · Portrait user studyaverage user score (1-5): 4.1 -> 3.8Mixed
replace Fast-Gatys2016 guided transfer net with the proposed methodportrait style transfer · Portrait user studyaverage user score (1-5): 1.5 -> 3.8Improved
add multi-layer style loss instead of single-layer style loss (ablation)proposed method · Qualitative closeups (Fig. 10)texture transferImproved
use relu4_1 instead of relu3_1 for decoder and fusion (ablation)proposed method · Qualitative (Fig. 8)color distributionHurt
use relu5_1 instead of relu3_1 (ablation)proposed method · Qualitative (Fig. 12)result qualityHurt
fuse features by iterative reconstruction instead of neural patch swap (ablation)proposed method · Qualitative (Fig. 11)texture transferImproved
increase feature-fusion iterations N beyond 500 (ablation)proposed method · Qualitative (Fig. 7)convergenceNo clear effect
add segmentation masks in feature spaceproposed method · Qualitativefailure cases such as ground textures on skyImproved

Closest papers by meaning

Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.

Open scime draft

Get the next scime Dispatch and a monthly field update

Free. We send a confirmation link first; every email has a one-click unsubscribe. Privacy