MUMBAI, India, June 26 -- Intellectual Property India has published a patent application (202641071789 A) filed by Sr University on June 10, 2026, for A Self-Calibrating Multimodal Sensor Fusion System For Pedestrian Detection With Uncertainty- Aware Deep Neural Networks.
Inventors include Deepa G; Dr. Johnson Kolluri; and Dr. Nagunuri. Rajender.
The application for the patent was published on June 19, 2026, under issue no. 25/2026.
Abstract: ABSTRACT Disclosed herein is a self-calibrating multimodal sensor fusion system (100) for pedestrian detection with uncertainty-aware deep neural networks, the system (100) comprises a sensor data acquisition module (102) configured to receive multimodal sensory inputs. The system also includes a feature extraction module (104) configured to generate modality-specific feature representations. The system also includes a confidence estimation module (106) configured to determine, for each sensory modality, a confidence distribution representative of uncertainty. The system also includes a Bayesian fusion module (108) configured to probabilistically combine the modality-specific feature representations. An online self-calibration module (110) configured to continuously monitor geometric drift and radiometric drift. The system also includes a pedestrian detection module (112) configured to process the fused pedestrian detection predictions. The system also includes an uncertainty- aware non-maximum suppression module (114) configured to evaluate confidence bounds. The system also includes an edge deployment optimization module (116) configured to perform structured neural network.
Disclaimer: Curated by HT Syndication.
