Abstract
While security camera footage can be searched by date/time ranges, it is not possible to search footage using descriptive text phrases. Consequently, investigations of security camera footage are laborious and time consuming. This disclosure describes techniques that enable security operators to search for individuals of interest in security camera footage with descriptive text queries in an efficient, privacy-sensitive manner. Per the techniques, images of individuals captured on security camera footage are optionally modified to blur facial features. After the blurring, embeddings are obtained for different individual images captured in the footage, where similar individuals cluster together in the embedding space. Using generative artificial intelligence, an artificial image is generated that matches a text phrase that describes a person of interest. Image generation can be guided by a user till the generated image captures the query attributes accurately. An embedding is generated for the artificial image and a search of matching embeddings is performed to identify people in the video footage.
Creative Commons License
This work is licensed under a Creative Commons Attribution 4.0 License.
Recommended Citation
NA, "Descriptive Search of Security Camera Footage via Generative Artificial Intelligence", Technical Disclosure Commons, ()
https://www.tdcommons.org/dpubs_series/9511