Using Content and Trace Feature Extractors for Exposing Fake Faces through Deep Learning
NeuroQuantology · 2023
In this procedure, we make an effort to address the effect of these variables on the identification of a photo from a face sketch using a cascaded static and dynamic local feature extraction approach.to ensure that the feature vectors created are based on the appropriate patches.Along with that approach, we include the closest neighbour technique for matching sketches and photos, which concatenates the feature vectors obtained from local static extraction.Following the extraction of the image's characteristics, the most comparable images are narrowed down based on their closes neighbours.Eventually, feature vectors from the local dynamic extraction approach are used to rematch these images.The closest neighbour technique is used to match the feature vectors.The primary goals of this method are to increase the detection system's precision as well as identification rates and feature stability.