Archive | 2019
Combined Advertising Sign Classifier
Abstract
The article describes the problem of classifying photographs of advertising signs of commercial establishments according to the type of services provided. The proposed solution is based on the sharing of textual and visual features. We provide a composite model that includes a text recognition module and an extractor of visual characteristics to improve classification accuracy. We achieve \\(F_1\\) of 0.24 exceeding strong baseline quality for 10%.