MUMBAI, India, Sept. 28 -- Intellectual Property India has published a patent application (202641112617 A) filed by Vallurupalli Nageswara Rao Vignana Jyothi Institute Of Engineering And Technology on September 19, 2026, for A Deep Learning-Based Framework For Imbalanced Image Dataset Optimization Using Adaptive Sampling And Data Augmentation Techniques.
Inventor includes B. Venkatesh Reddy.
The application for the patent was published on September 25, 2026, under issue no. 39/2026.
Abstract: ABSTRACT [0019] A method is presented for correcting severe class imbalance in medical image collections through the controlled generation of synthetic samples. A conditional generative adversarial network is trained on a skin-disease dataset containing nine categories and only nine hundred images in total. Some disease classes originally contain as few as five to ten examples while others exceed eighty. The generator produces new images conditioned on class labels, and the discriminator evaluates their authenticity. After training, synthetic images are added only to the under-represented classes until the overall distribution becomes balanced. The resulting dataset is visualized with bar graphs and made available through an interactive interface. Experiments confirm that the balanced collection improves training stability and reduces bias toward majority classes, offering a practical solution for medical image classification tasks where minority-class data are scarce.
Disclaimer: Curated by HT Syndication.
