AI Paper Atlas

Research map / Label-efficient and robust learning / Self-supervised and contrastive

Rethinking Vision Transformer and Masked Autoencoder in Multimodal Face Anti-Spoofing

Zitong Yu, Rizhao Cai, Yawen Cui, Xin Liu, Yongjian Hu, Alex Chichung Kot

IJCV 2024 · Label-efficient and robust learning › Self-supervised and contrastive · approach: Masked autoencoder pretraining · 83 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
add AMA adapter finetuning instead of direct finetuningViT-Base · WMCA unseen, RGB+IR+DepthACER mean (%): 6.46 -> 5.56Improved
add AMA adapter finetuningViT-Base · MmFA RGB+DepthACER (%): 4.60 -> 1.84Improved
add ConvAdapter finetuningViT-Base · MmFA RGBACER (%): 17.37 -> 10.02Improved
add M2A2E pre-training to AMAViT+AMA · WMCA unseen, RGB+IR+DepthACER mean (%): 5.56 -> 3.86Improved
add M2A2E pre-training to ConvAViT+ConvA · WMCA unseen, RGBACER mean (%): 9.42 -> 11.5Hurt
add M2A2E pre-training to AMAViT+AMA · WMCA unseen, RGB+DepthACER mean (%): 5.78 -> 5.82No clear effect
add M2A2E pre-training to AMAViT+AMA · Cross-dataset WMCA->MmFA, RGB+IR+DepthHTER (%): 13.99 -> 8.60Improved
add M2A2E pre-training to ConvAViT+ConvA · MmFA intra-test, RGBACER (%): 10.02 -> 6.40Improved
add M2A2E pre-training to ConvAViT+ConvA · Cross-dataset WMCA->MmFA, RGBHTER (%): 21.31 -> 23.56Hurt
add adaptive modality weighting (AMA) over multimodal ConvAdapter (ablation)multimodal ConvAdapter · WMCA seen, multimodalACER (%): decreases 0.52%, 0.31%, 1.07%, 0.07%Improved
use multimodal ConvAdapter instead of independent ConvAdapter (ablation)ConvAdapter · WMCA seen, multimodalACER (%): reduces more than 0.5% ACERImproved
use huge multimodal ConvAdapter (D'xK output) (ablation)multimodal ConvAdapter · WMCA seen, multimodalACER (%)No clear effect

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