MUMBAI, India, Sept. 28 -- Intellectual Property India has published a patent application (202641112670 A) filed by Essae Teraoka Pvt Ltd on September 20, 2026, for An Edge-Deployed Item Recognition Apparatus And Method For Open-Set Recognition Of Items.
Inventors include Aakarsh Prabhu; Nagarajan Vishwanathan; Anand Prabhu; Bargava Subramanian; Sandhiya Cv; and Raghotham Sripadraj.
The application for the patent was published on September 25, 2026, under issue no. 39/2026.
Abstract: The present invention relates to an edge-deployed item-recognition apparatus (106) and methods (700, 800) for open-set recognition of items. The apparatus generates an embedding vector from a captured image and compares the embedding vector with registered embedding vectors stored in a registration/embedding database (108, 210, 412). A trained embedding generation unit (218, 404) is trained using supervised contrastive loss, wherein objects not belonging to recognised item categories are trained as an additional class and positioned within a distinguishable region of an embedding space. An open-set rejection unit (222, 416) withholds a recognition output when the generated embedding corresponds to the additional class and otherwise provides a recognition result and confidence score (110, 418) based on computed similarity. The invention further enables registration of new item categories by storing corresponding embedding vectors in the registration/embedding database (108, 210, 412) without modifying or retraining the embedding generation unit (218, 404), thereby enabling subsequent recognition of newly registered items.
Disclaimer: Curated by HT Syndication.
