MUMBAI, India, Oct. 26 -- Intellectual Property India has published a patent application (202631088139 A) filed by Aryan Singh; Hari Om; Ankit Singh; Avdhesh Kumar; and Manish Pratap Singh on July 20, 2026, for Lightweight Spatio-Temporal Modeling For Real-Time Violence Detection In Surveillance Videos.
Inventors include Aryan Singh; Hari Om; Ankit Singh; Avdhesh Kumar; and Manish Pratap Singh.
The application for the patent was published on October 23, 2026, under issue no. 43/2026.
Abstract: The present invention relates to a lightweight spatio-temporal modeling system for real-time violence detection in surveillance video feeds. The system comprises an input preprocessing module configured to receive surveillance videos and generate a five-channel input tensor including three RGB channels and two optical flow channels. A dual-stream neural network independently processes the RGB and optical flow inputs through respective processing streams for extracting spatial and motion features. One or more parameter-free Temporal Shift Modules (TSMs) are embedded within each processing stream to facilitate temporal feature modeling by exchanging feature information across adjacent frames without increasing trainable parameters. The extracted features are combined using an element-wise multiplicative fusion module to generate fused spatio-temporal representations, which are classified by a classification module to identify violent or non-violent activities. The system enables computationally efficient spatio-temporal feature extraction suitable for real-time deployment in intelligent video surveillance applications.
Disclaimer: Curated by HT Syndication.
