The article describes a software pipeline for identifying and classifying the interests of users in social networks using modern models and deep learning methods. The developed program is able to detect the presence of bad habits (smoking, alcohol), a sporting lifestyle, as well as determine the user's addiction to travel by an available set of photos. The software includes modules that implement deep learning algorithms for the object detection and semantic segmentation of images using the Cascade-R-CNN and DeepLabv3+ models, and the module for converting annotations of the images from COCO, ImageNet, OpenImagesV6 datasets and manually labeled images to the unified format. The models were trained on the created original datasets which include 90200 photos in total. The accuracy of the developed models is from 83.7% up to 86.6% mAP for object detection depending on a specific category of objects and 78.4% pixel accuracy for segmentation.
scite is a Brooklyn-based organization that helps researchers better discover and understand research articles through Smart Citations–citations that display the context of the citation and describe whether the article provides supporting or contrasting evidence. scite is used by students and researchers from around the world and is funded in part by the National Science Foundation and the National Institute on Drug Abuse of the National Institutes of Health.