1 paper · 1 filter
Jiaming Liang, Yuwan Xue, Haowei Liu +9
In existing splicing forgery datasets, the insufficient semantic variety of spliced regions causes trained detection models to overfit semantic features rather than learn genuine s…