Dvmm 191 Jun 2026

Developed by the Digital Video and Multimedia Lab (DVMM) at Columbia University, this dataset serves as a benchmark for researchers developing "blind" image forgery detection systems. Dataset Composition

Maya realized naming was the pivot. With nomenclature came narrative, and with narrative came vulnerability. To stitch a name onto fragments invited attachment. To keep names anonymous was a kind of mercy—but also a denial of personhood. dvmm 191