AI Paper Atlas

Research map / Architectures and efficiency / Attention and transformers

TransForensics: Image Forgery Localization With Dense Self-Attention

Jing Hao, Zhixin Zhang, Shicai Yang, Di Xie, Shiliang Pu

ICCV 2021 · Architectures and efficiency › Attention and transformers · approach: Efficient image transformers · 112 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 self-attention and positional encoding to FCN (ablation)FCN (ResNet50) · CASIAAUC: 0.669 -> 0.809Improved
add self-attention and positional encoding to FCN (ablation)FCN (ResNet50) · COVERAGEAUC: 0.715 -> 0.862Improved
add self-attention and positional encoding to FCN (ablation)FCN (ResNet50) · IMD2020AUC: 0.735 -> 0.815Improved
add self-attention without positional encoding (ablation)FCN (ResNet50) · CASIAAUC: 0.669 -> 0.655Hurt
add positional encoding without self-attention (ablation)FCN (ResNet50) · COVERAGEAUC: 0.715 -> 0.707Hurt
add dense correction (ablation)self-attention + positional encoding model · CASIAAUC: 0.809 -> 0.837Improved
add dense correction (ablation)self-attention + positional encoding model · COVERAGEAUC: 0.862 -> 0.883Improved
add dense correction (ablation)self-attention + positional encoding model · IMD2020AUC: 0.815 -> 0.848Improved
fuse branches by multiplication instead of addition (ablation)TransForensics · CASIAAUC: 0.739 -> 0.837Improved
fuse branches by multiplication instead of addition (ablation)TransForensics · COVERAGEAUC: 0.857 -> 0.883Improved
fuse branches by multiplication instead of addition (ablation)TransForensics · IMD2020AUC: 0.827 -> 0.848Improved
use low-level C2 branch rather than deep C5 branch as output (ablation)TransForensics · IMD2020AUC: 0.827 -> 0.848Improved

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