The article discusses a case study on annotating real estate images using a semi-supervised learning approach. The authors propose a model for automatic pre-selection of images from large unlabeled datasets for training ConvNets. The study aims to investigate how the proposed model processes complex properties of real estate photographs and which domain-specific labels are generalized well by the classifiers. The challenges of this task include varying image resolutions, lighting conditions, and a limited number of relevant labels.
Publication date: 27 Sep 2023
Project Page: Not provided
Paper: https://arxiv.org/pdf/2309.15097
