Use of Machine Learning Classification Models, both Image and Text, in the Network Graphing. Case Study: Community of Practice Among Graffiti Writers on Freight Trains

Lecture presented at the 3rd International Digital Humanities Conference (DH_BUDAPEST_2022) in Budapest, Hungary. Describes a technical architecture combining computer vision and natural language processing to structure graph-oriented relational databases.

The methodology proposes using CNN convolutional models trained in Roboflow to segment and categorize graffiti styles, and spaCy NLP pipelines to analyze the identity vocabulary of textual descriptions. The result is plotted as a complex hypertextual network that correlates the visual aesthetics of graffiti with the semantic patterns and subcultural affinity of writers.