General purpose and interactive video analytics

Stanford University, 2023


Abstract

This dissertation develops systems that make large video collections easier, faster, and more cost-effective to query, including heterogeneous serverless execution, declarative query optimization, and vision-language-model search.