Content Bias in Deep Learning Image Age Approximation: A new Approach Towards better Explainability

arXiv:2310.02067v3 Announce Type: replace
Abstract: In the context of temporal image forensics, it is not evident that a neural network, trained on images from different time-slots (classes), exploits solely image age related features. Usually, images taken in close temporal proximity (e.g., belonging to the same age class) share some common content properties. Such content bias can be exploited by a neural network. In this work, a novel approach is proposed that evaluates the influence of image content. This approach is …

abstract age approximation arxiv bias context cs.ai cs.cv deep learning explainability exploits features forensics image images network neural network temporal type