AI Paper Atlas

Research map / Video / Video generation

Speech-driven Facial Animation using Cascaded GANs for Learning of Motion and Texture

Dipanjan Das, Sandika Biswas, Sanjana Sinha, Brojeshwar Bhowmick

ECCV 2020 · Video › Video generation · approach: Talking face generation · 105 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
use adaptive instance normalization instead of channel-wise concatenation (ablation)t-GAN texture generator · TCD-TIMIT texture ablationPSNR: 27.2 -> 28.3Improved
use adaptive instance normalization instead of channel-wise concatenation (ablation)t-GAN texture generator · TCD-TIMIT texture ablationLMD: 1.65 -> 1.57Improved
add feature loss (ablation)t-GAN texture generator · TCD-TIMIT texture ablationPSNR: 28.3 -> 28.9Improved
add adversarial loss (ablation)t-GAN texture generator · TCD-TIMIT texture ablationPSNR: 28.9 -> 30.7Improved
add adversarial loss (ablation)t-GAN texture generator · TCD-TIMIT texture ablationLMD: 1.5 -> 1.4Improved
add discriminator (GAN) to landmark generator l-GAN (ablation)l-GAN · TCD-TIMITqualitative lip motionImproved
use context window of 6 frames for DeepSpeech features (ablation)l-GAN · TCD-TIMITLMDImproved
add L2 regularization loss to the motion loss (ablation)l-GAN · TCD-TIMITLMDImproved
add direction loss to the motion loss (ablation)l-GAN · TCD-TIMITconvergence speedImproved
add synthetic audio noise up to -30dB (ablation)l-GAN lip sync · TCD-TIMITLMDNo clear effect
fine-tune for 10 epochs instead of 1 epoch (ablation)t-GAN · GRIDPSNR: 21.5 -> 29.8Improved
fine-tune for 10 epochs instead of 1 epoch (ablation)t-GAN · TCD-TIMITPSNR: 20.6 -> 30.7Improved

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