AI Paper Atlas

Research map / Detection and segmentation / Open-set and few-shot detection

ObjectFormer for Image Manipulation Detection and Localization

Junke Wang, Zuxuan Wu, Jingjing Chen, Xintong Han, Abhinav Shrivastava, Ser-Nam Lim, Yu-Gang Jiang

CVPR 2022 · Detection and segmentation › Open-set and few-shot detection · approach: Open-set and unknown object detection · 225 citations (OpenAlex)

Check 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 ObjectFormer instead of PSCCNetImage manipulation localization · Coverage, fine-tunedlocalization AUC (%): 94.1 -> 95.7Improved
Use ObjectFormer instead of PSCCNetImage manipulation localization · CASIA, fine-tunedlocalization AUC (%): 87.5 -> 88.2Improved
Use ObjectFormer instead of PSCCNetImage manipulation localization · NIST16, fine-tunedlocalization AUC (%): 99.6 -> 99.6No clear effect
Use ObjectFormer instead of PSCCNetImage manipulation detection · CASIA-Ddetection AUC (%): 99.65 -> 99.70Improved
Add high-frequency feature extraction (ablation)ObjectFormer · CASIAlocalization AUC (%): 14.6%Improved
Add high-frequency feature extraction (ablation)ObjectFormer · NIST16localization AUC (%): 11.0%Improved
Add boundary-sensitive contextual incoherence modeling (BCIM) (ablation)ObjectFormer · CASIAlocalization AUC (%): 6.2%Improved
Add boundary-sensitive contextual incoherence modeling (BCIM) (ablation)ObjectFormer · NIST16localization AUC (%): 2.4%Improved
Use object prototypes and stacked object encoders/patch decoders rather than vanilla self-attention blocks (ablation)ObjectFormer · NIST16localization AUC (%): 5%Improved
Use object prototypes and stacked object encoders/patch decoders rather than vanilla self-attention blocks (ablation)ObjectFormer · NIST16localization F1 (%): 12.5%Improved
Increase the number of object prototypes, with the best reported setting at N=16 (ablation)ObjectFormer · Columbia and CASIAlocalization performanceImproved

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