Problem
Research questions and friction points this paper is trying to address.
Image Forgery Localization
Generalization
Benchmark
Cross-dataset Evaluation
Distribution Shift
GIFTBench: Diagnosing Generalization in Image Forgery Localization and Informing Model Design
🤖 AI Summary
This study addresses the limited coverage and entangled factors in existing image forgery localization benchmarks, which hinder comprehensive evaluation of model generalization. To this end, we construct GIFTBench, a multi-axis disentangled benchmark that reveals asymmetric cross-domain transfer characteristics. We further propose ForenScope, a framework that integrates pixel-level annotations, multi-depth multi-scale features, and a selective coarse-grained conditioning mechanism to enhance localization performance. Experimental results demonstrate that our approach significantly improves cross-dataset localization accuracy while preserving image-level detection capabilities. These findings validate the dual value of GIFTBench as both a diagnostic tool and a large-scale training resource for advancing robust image forgery analysis.
📝 Abstract
Reliable evaluation of image forgery localization (IFL) requires assessing models under diverse distribution changes, yet existing benchmarks often cover limited manipulation conditions or entangle multiple factors in cross-dataset evaluation. Consequently, aggregate performance provides an incomplete view of localization generalization. We introduce GIFTBench, a multi-axis benchmark of 115,013 manipulated images with pixel-level annotations spanning manipulation source, semantic target, editing operation, and composition complexity. GIFTBench supports axis-specific transfer analysis and evaluation on twelve external datasets. Its diagnostic studies reveal asymmetric cross-source transfer, recall-dominated failures, and heterogeneous degradation across semantic, operational, and compositional changes. Beyond diagnosis, the scale and diversity of GIFTBench provide a substantially broader training distribution than conventional IFL datasets. Training representative localizers on GIFTBench consistently improves their aggregate transfer to external datasets, showing that the benchmark serves not only as an evaluation tool but also as an effective training resource for cross-domain localization. Guided by the diagnostic findings, we further develop ForenScope, a detection and localization framework combining classification-adapted representations with multi-depth, multi-scale spatial features, learned layer fusion, and selective coarse-scale conditioning. Experiments show improved cross-dataset localization while retaining image-level detection capability. The GIFTBench dataset showcase page is available at https://giftbench-preview.doudoudouya337.chatgpt.site.
Innovation
Methods, ideas, or system contributions that make the work stand out.
Image Forgery Localization
Multi-axis Benchmark
Cross-domain Generalization
ForenScope
Learned Layer Fusion
🔎 Similar Papers
No similar papers found.
