AI Paper Atlas

Research map / Image generation / Editing and style transfer

Learned Spatial Representations for Few-Shot Talking-Head Synthesis

Moustafa Meshry, Saksham Suri, Larry S. Davis, Abhinav Shrivastava

ICCV 2021 · Image generation › Editing and style transfer · approach: Texture and sketch synthesis · 46 citations (OpenAlex)

Read the paperCheck your idea against 137,217 papers

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 UNet generator with SPADE conditioned on landmarks (ablation)FSTH-style landmark-driven generator · VoxCeleb ablationFID: 67.19 -> 56.00Improved
replace UNet generator with SPADE conditioned on landmarks (ablation)FSTH-style landmark-driven generator · VoxCeleb ablationLPIPS: 0.274 -> 0.268No clear effect
add learned segmentation maps as SPADE input (ablation)FSTH-style landmark-driven generator + SPADE · VoxCeleb ablationidentity similarity: 0.834 -> 0.828Hurt
add learned segmentation maps as SPADE input (ablation)FSTH-style landmark-driven generator + SPADE · VoxCeleb ablationFID: 56.00 -> 62.78Hurt
add learned segmentation maps as SPADE input (ablation)FSTH-style landmark-driven generator + SPADE · VoxCeleb ablationLPIPS: 0.268 -> 0.265Improved
replace supervised segmentation with learned latent spatial layout (ablation)FSTH-style landmark-driven generator + learned seg. maps · VoxCeleb ablationLPIPS: 0.265 -> 0.247Improved
replace supervised segmentation with learned latent spatial layout (ablation)FSTH-style landmark-driven generator + learned seg. maps · VoxCeleb ablationFID: 62.78 -> 54.40Improved
replace supervised segmentation with learned latent spatial layout (ablation)FSTH-style landmark-driven generator + learned seg. maps · VoxCeleb ablationidentity similarity: 0.828 -> 0.860Improved
use oracle ground-truth segmentations as spatial input (upper bound) (ablation)SPADE generator · VoxCeleb ablationFID: 54.40 -> 48.06Improved
use proposed method instead of FOMMtalking-head synthesis · VoxCeleb single-shotPSNR: 18.20 -> 17.37Hurt
use proposed method instead of FOMMtalking-head synthesis · VoxCeleb single-shotidentity similarity: 0.869 -> 0.886Improved
use proposed method instead of FSTHtalking-head synthesis · VoxCeleb single-shotFID: 53.07 -> 45.69Improved

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