English (United Kingdom)

https://curated-unify.zendy.io/wp-json/zendy-region/v1/featured_content/oa?rat=en

https://curated-unify.zendy.io/wp-json/zendy-region/v1/highlighted_journal/

Global: Zendy Plus annual

Presents the access of premium content as premium feature

Premium Content

Presents the keyphrase highlighting as premium feature

Keyphrase Highlighting

Presents the summarisation as premium feature

Summarisation

Insights

Presents the pdf analysis as premium feature

PDF Analysis

Presents the zaia usage as premium feature

ZAIA

Global: Zendy Plus monthly

Global: Zendy Tools annual

Global: Zendy Tools monthly

Global: Zendy Free Plan

Despite significant advancements of deep learning-based forgery detectors fordistinguishing manipulated deepfake images, most detection approaches sufferfrom moderate to significant performance degradation with low-qualitycompressed deepfake images. Because of the limited information in low-qualityimages, detecting low-quality deepfake remains an important challenge. In thiswork, we apply frequency domain learning and optimal transport theory inknowledge distillation (KD) to specifically improve the detection oflow-quality compressed deepfake images. We explore transfer learning capabilityin KD to enable a student network to learn discriminative features fromlow-quality images effectively. In particular, we propose the Attention-basedDeepfake detection Distiller (ADD), which consists of two novel distillations:1) frequency attention distillation that effectively retrieves the removedhigh-frequency components in the student network, and 2) multi-view attentiondistillation that creates multiple attention vectors by slicing the teacher'sand student's tensors under different views to transfer the teacher tensor'sdistribution to the student more efficiently. Our extensive experimentalresults demonstrate that our approach outperforms state-of-the-art baselines indetecting low-quality compressed deepfake images.

ADD: Frequency Attention and Multi-View based Knowledge Distillation to Detect Low-Quality Compressed Deepfake Images