Browse Teach Curate

Deep Style Extrapolation for Artistic-Style Classification of Image Fragments

A deep-learning framework that predicts the artistic style of a partial, irregularly shaped image fragment by first passing it through a 'style extrapolation' module - a modified convolutional autoencoder trained with custom style and content losses that amplifies stylistic cues while preserving content - and then classifying the transformed fragment with a transfer-learning image classifier, robust to the number of candidate styles and to fragment geometry.

Rosie Text

Recognizing Artistic Style of Archaeological Image Fragments Using Deep Style Extrapolation Gur

Archaeological artworks such as Pompeian wall frescoes survive only as mixed, eroded fragments, and sorting the pieces by artistic style is a hard computer-vision classification problem because each …