Intellectual Property India Publishes Patent Application for Adaptive Machine Learning And Deep Learning-Based System For Sketck-To-3d Reconstruction In Forensic Application

MUMBAI, India, Oct. 5 -- Intellectual Property India has published a patent application (202641115650 A) filed by Rns Institute Of Technology, Bengaluru.; Ms. Vinutha G Kassistant Professor; Dr. Vidya Yassociate Professor; Deekshitha B Rrncs; Deepthi A Srncs; Dhanushri Lrncs; and Jayashree S Mrncs on September 27, 2026, for Adaptive Machine Learning And Deep Learning-Based System For Sketck-To-3d Reconstruction In Forensic Application.

Inventors include Ms. Vinutha G Kassistant Professor; Dr. Vidya Yassociate Professor; Deekshitha B Rrncs; Deepthi A Srncs; Jayashree S Mrncs; and Dhanushri Lrncs.

The application for the patent was published on October 02, 2026, under issue no. 40/2026.

Abstract: Forensic investigation systems often depend on sketches created by witnesses when no real facial images are available. However, these sketches usually lack detailed facial features, which lowers identification accuracy. Traditional methods involve manually comparing sketches with existing databases, a process that takes a lot of time and is susceptible to human error. The proposed system offers an automated way to reconstruct faces from forensic sketches using a blend of Machine Learning and Deep Learning techniques. It uses a dataset with pairs of sketches and facial images, which go through preprocessing steps like normalization, noise reduction, and feature extraction. Various Deep Learning models, including Convolutional Neural Networks (CNN), ResNet, Generative Adversarial Networks (GAN), and Autoencoders, are trained and assessed using metrics such as accuracy, precision, recall, and loss. Additionally, Machine Learning models like Support Vector Machine (SVM), Random Forest, and XGBoost are applied for feature analysis and comparison.An adaptive model selection method helps find the best model based on these performance metrics. The chosen model creates a high-quality 2D facial image, which is then converted into a 3D facial representation using a pre- trained 3D Morphable Model. This system improves reconstruction accuracy, minimizes manual work, and serves as an efficient omputational tool for forensic identification and analysis. The proposed system enhances reconstruction accuracy, reduces manual intervention, and provides an efficient computational tool for forensic identification and analysis.

Disclaimer: Curated by HT Syndication.