MUMBAI, India, Nov. 2 -- Intellectual Property India has published a patent application (202641125348 A) filed by Vellore Institute Of Technology on October 21, 2026, for System And Method For Aerial Scene Classification By Input- Conditioned Class-Wise Fusion.
Inventors include Balaji Gn; and Bhumi Rana.
The application for the patent was published on October 30, 2026, under issue no. 44/2026.
Abstract: Thepresent invention discloses asystem and method forfine-grained aerial sceneclassification by class-wise fusion of classical and quantum decisions. A pre-processed aerial image (104) is passed through a Swin Transformer backbone (106) and an embedding head (108) to a 128- dimensional embedding. A classical head (110) produces classical logits. A quantum adapter (112) reduces the embedding to 12 values, which a 12-qubit variational quantum circuit (114) encodes with data re- uploading and strongly entangling layers, and a quantum head (116) maps the Pauli-Z expectations to quantum logits. A gate network (118) computes from the embedding one sigmoid gate value per class, and a fusion unit (120) forms each final class logit as the gate-weighted sum of the classical and quantum logits of that class. On the 45-class NWPU-RESISC45 dataset the system reached 97.25 percent test accuracy.
Disclaimer: Curated by HT Syndication.
