Hidden Layers LLC: eCommerce via Picture Search

Abstract

Computer Vision, a subdivision of Artificial Intelligence, has improved rapidly, especially since 2015, when Google made open-source the language most fit for the task, TensorFlow. A computer’s ability to “see” is an absolute pre-requisite, among other uses cases, for autonomous vehicles. Such technologies are already being deployed on our streets. e-commerce has embraced Artificial Intelligence as well. The most readily-observable implementation is the ability simply to type words into a site and be offered relevant, if not exact-match, desired products. User interface via speech, such as Alexa and Siri, provide second modalities of human-computer interaction via Artificial Intelligence. Finally, many sites are now equipped with recommender systems, which employ algorithms to consider other product offerings that the consumer may want. What has not yet become part of the mix, including for larger and more able sites, is search by image. In this thesis, the author explains his solution, hidden layers LLC. hidden layers allows a consumer to upload a picture and be offered the most likely matches of the desired product through Computer Vision.