At idealo.de (a leading price comparison website in Europe), we have a dedicated service to provide hotel price comparisons (hotel.idealo.de). For each hotel, we receive dozens of images and face the challenge of composing image galleries that are attractive and at the same time help our users to make informed decisions. Given that we have millions of hotel offers, we end up with more than 100 million images for which we need both an attractiveness assessment and a tag (e.g. a “bathroom” or “bedroom” tag).
We addressed the need to automatically assess image quality by implementing an aesthetic and technical image quality classifier based on Google’s research paper “NIMA: Neural Image Assessment”. NIMA consists of two Convolutional Neural Networks (CNN) that aim to predict the aesthetic and technical quality of images, respectively. Additionally, another CNN was trained to assign hotel area tags, like bathroom, bedroom, pool, etc. to each image. The models were trained via transfer learning, where ImageNet pre-trained CNNs were fine-tuned for each classification task. In this talk, we will present the insights that we’ve gained throughout the process and discuss the challenges we faced while putting such a system in production.