The article introduces PhotoScout, a tool designed to perform semantic image search tasks. Unlike existing tools that rely on metadata or visual similarity, PhotoScout allows users to provide natural language descriptions, positive and negative examples, and object tags to specify their search tasks. The tool is powered by a program synthesis engine that generates visual queries and executes the synthesized program to retrieve the desired images. The study found that PhotoScout improved the accuracy of image retrieval tasks and reduced manual effort for users.
Publication date: 19 Jan 2024
Project Page: https://arxiv.org/abs/2401.10464v1
Paper: https://arxiv.org/pdf/2401.10464
