MUMBAI, India, June 26 -- Intellectual Property India has published a patent application (202621053232 A) filed by Dr. D. Y. Patil Institute Of Technology, Pimpri, Pune - on April 27, 2026, for Botanica – Utilizing Deep Learning For Precise Identification Of Medicinal Plants And Leaves.
Inventors include Ms. Arti Vilas Sonawane; Gayatri Bhandge; Maansi Sonawane; Vishvachaitanya Wagh; and Tejas More.
The application for the patent was published on June 19, 2026, under issue no. 25/2026.
Abstract: This work further explores data augmentation techniques, including rotation, scaling, and illumination normalization, to improve model generalization across diverse field conditions. A curated dataset of high-resolution leaf images representing multiple Ayurvedic species is compiled, annotated, and split into training, validation, and testing sets. Transfer learning with pre-trained ResNet-101V2 weights accelerates convergence while reducing overfitting. Hyperparameters, such as learning rate, batch size, and optimizer choice, are systematically tuned using cross-validation. Class activation maps are employed to interpret model decisions and highlight discriminative leaf regions. The system is evaluated using accuracy, precision, recall, and F1-score, demonstrating robust performance under varying backgrounds and lighting conditions. A lightweight deployment pipeline is proposed for mobile and edge devices, enabling real-time identification in field settings. Comparative analysis confirms advantages over baseline models, while ablation studies validate component contributions. The approach supports scalable biodiversity monitoring, assists practitioners in accurate plant selection, and reduces manual effort. Future work will expand species coverage, incorporate multi-modal data, and explore few-shot learning for rare classes. Additionally integration with geotagging and cloud services will enhance accessibility.
Disclaimer: Curated by HT Syndication.
