1 paper
Juho Jung, Sangyoun Lee, Jooeon Kang +1
All current benchmarks for multimodal deepfake detection manipulate entire frames using various generation techniques, resulting in oversaturated detection accuracies exceeding 94%…