User-Centered Visual Search

This post details the evolution of Palantir's User-Centered Machine Learning (UCML) workflow by introducing Visual Search. This capability allows end-users to independently and in real-time initiate bespoke object detection needs on demand, without requiring pre-trained models or data labeling. Users delineate pixels of interest and can optionally provide text guidance, which foundational vision-language models use to extract features and identify similar objects across images. This enables analysts to generate their own object detection capability for any object of interest, in any image, in real-time, without data labeling or model training. These detections can be fused with Ontology data for enriched analysis.