AI Paper Atlas

Research map / Detection and segmentation / CNN features

MVSS-Net: Multi-View Multi-Scale Supervised Networks for Image Manipulation Detection

Chengbo Dong, Xinru Chen, Ruohan Hu, Juan Cao, Xirong Li

TPAMI 2022 · Detection and segmentation › CNN features · approach: Forgery detection · 409 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
add image classification loss (ablation)Seg (FCN-16 with DA) · DEF-12k (trained on DEF-84k)pixel-level mean F1: 54.6 -> 46.3 pixel F1; image F1 70.9 -> 77.3Mixed
add noise-sensitive branch (NSB) (ablation)Seg+Clf · DEF-12k (trained on DEF-84k)pixel-level mean F1: 46.3 -> 50.8Improved
add edge-supervised branch (ESB) (ablation)Seg+Clf · DEF-12k (trained on DEF-84k)pixel-level mean F1: 46.3 -> 51.8Improved
use ESB instead of the edge branch of GSR-Net (ablation)Seg+Clf · DEF-12k (trained on DEF-84k)pixel-level mean F1: 49.9 -> 51.8Improved
add Sobel operation to ESB (ablation)Seg+Clf+Edge · DEF-12k (trained on DEF-84k)pixel-level mean F1: 50.5 -> 51.8; copy-move 0.382 -> 0.405Improved
fuse ESB and NSB with dual attention (ablation)Seg+Clf+Edge · DEF-12k (trained on DEF-84k)Com-F1: 62.6 -> 64.3Improved
dual attention fusion instead of model-averaging ensemble (ablation)Seg+Clf+Noise and Seg+Clf+Edge · DEF-12k (trained on DEF-84k)Com-F1: 62.2 -> 64.3Improved
use ConvGeM instead of GMP as Gclf (ablation)MVSS-Net · DEF-12k (trained on DEF-84k)Com-F1: 64.3 -> 66.3Improved
use GeM instead of GMP as Gclf (ablation)MVSS-Net · DEF-12k (trained on DEF-84k)Com-F1: 64.3 -> 64.3; pixel F1 53.8 -> 56.2, image sensitivity 79.7 -> 63.1No clear effect
edge segmentation head instead of bbox regression head (ablation)MVSS-Net++ · DEF-12k (trained on DEF-84k)Com-F1: 65.7 -> 66.3Improved
add median filtering residual (MFR) blocks to NSB (ablation)MVSS-Net++ · DEF-12k (trained on DEF-84k)Com-F1: 66.3 -> 65.1; image sensitivity 74.8 -> 81.3, pixel F1 56.7 -> 54.6Hurt
add dual attention (DA) to FCN-16 backbone (ablation)FCN-16 · DEF-12k (trained on DEF-84k)Com-F1: 45.5 -> 61.7Improved

Later papers that compared with this method

3 later accepted papers report MVSS-Net as a baseline in their results tables; 3 of them report a better number on most of the results they share with it.

Earlier methods this paper compares with

Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 1 on most shared results.

Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.

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