MUMBAI, India, Aug. 12 -- Intellectual Property India has published a patent application (202531125451 A) filed by Indian Institute Of Technology Patna on December 11, 2025, for System And Method For Remote Sensing Visual Question Answering Using Multi-Modal Transformer Fusion.
Inventors include Suman Kumar Maji; and Anirban Saha.
The application for the patent was published on August 07, 2026, under issue no. 32/2026.
Abstract: The invention relates to a system (100) and a computer-implemented method for remote sensing visual question answering. The system includes an input module (108) for receiving a remote sensing image and a natural language question. A representation module (102) comprises a visual representation module (104) having encoder blocks (120), downsampler blocks (118), and decoder blocks (116) to generate multi-scale visual feature representations, and a question representation module (106) using an LLM based feature extractor (122) and a transformer language unit (124) to generate a contextual embedding. A category classifier (126) classifies the contextual embedding into predefined categories. A multi-layer correlated attention driven feature fusion module (110) performs cross-attention between visual and textual features to obtain a fused multimodal representation. A category-aware answering module (112) conditions the fused representation with the predicted category to generate an answer, which is provided through an output module (114). Refer Fig. 1
Disclaimer: Curated by HT Syndication.
