A Novel Framework for Assessment of Learning-based Detectors in Realistic Conditions with Application to Deepfake Detection
This work addresses the gap in assessing detectors for real-world applications like deepfake detection, though it is incremental as it builds on existing methods with a new evaluation framework.
The paper tackles the problem of evaluating learning-based detectors under realistic conditions, focusing on deepfake detection, and shows that a data augmentation strategy based on natural image degradation significantly improves generalization, with unspecified but implied performance gains.
Deep convolutional neural networks have shown remarkable results on multiple detection tasks. Despite the significant progress, the performance of such detectors are often assessed in public benchmarks under non-realistic conditions. Specifically, impact of conventional distortions and processing operations such as compression, noise, and enhancement are not sufficiently studied. This paper proposes a rigorous framework to assess performance of learning-based detectors in more realistic situations. An illustrative example is shown under deepfake detection context. Inspired by the assessment results, a data augmentation strategy based on natural image degradation process is designed, which significantly improves the generalization ability of two deepfake detectors.