MUMBAI, India, April 17 -- Intellectual Property India has published a patent application (202641043720 A) filed by KKR & KSR Institute of Technology and Sciences, Guntur, Andhra Pradesh, on April 6, for 'a deep learning-driven classroom attendance system with explainable ai and incremental facial embedding updates.'
Inventor(s) include Dr. G. Murali; Ms. C. Parimala Tejaswi; Ms. Y. Hemalatha; Ms. P. Mounika; and Ms. M. Mahalakshmi.
The application for the patent was published on April 17, under issue no. 16/2026.
According to the abstract released by the Intellectual Property India: "This invention proposes a sophisticated, non-intrusive automated attendance management system that overcomes the inefficiencies of manual roll calls and the "black-box" nature of traditional biometric systems. The system architecture is built upon a high-performance deep learning pipeline optimized for crowded classroom environments. For the initial stage of image processing, the system implements YuNet, a high-speed, lightweight face detection model specifically chosen for its ability to handle varied scales and dense spatial distributions of faces in a single high-resolution classroom image.Following detection, facial feature extraction is performed using ArcNet (ArcFace), which utilizes an additive angular margin loss function to maximize the decision boundary between student identities in a high-dimensional hypersphere. This ensures that even with low-quality captures or varied student poses, the generated facial embeddings remain highly discriminative. These embeddings are compared against a pre-registered database using cosine similarity metrics to trigger automatic attendance logging. [0017] A primary novelty of this invention is the integration of Explainable AI (XAI) modules, which generate interpretable feedback such as confidence heatmaps and similarity distribution charts. This allows faculty to understand the rationale behind the AI's identification, specifically in cases of low-confidence matches, thereby increasing system transparency. Furthermore, the invention features Adaptive Vision capabilities, which utilize an incremental learning feedback loop to update stored student embeddings as their physical appearances evolve over a semester. Finally, the system includes a Learning Analytics engine that correlates attendance frequency with academic engagement trends, providing educators with actionable data on student participation. The resulting system ensures high throughput, prevents proxy attendance, and maintains student privacy by utilizing point-in-time image captures rather than continuous video surveillance."
Disclaimer: Curated by HT Syndication.
