Developed an automated system for segmenting ovarian follicles in ultrasound images to assist in diagnosing infertility, PCOS, and ovarian cancer. Using a 3D ultrasound dataset converted into 3,419 2D slices, we trained an Attention Residual U-Net model, which outperformed baseline U-Net and Attention U-Net models. Our model achieved a dice coefficient of 35.4%, accuracy of 91.25%, recall of 87.45%, precision of 91.13%, and IoU of 21.89%, demonstrating its effectiveness for medical image segmentation.

The proposed system is designed to monitor face mask compliance in crowded public areas using machine learning (ML), a subset of artificial intelligence (AI). Initially, data preprocessing was performed using the HAAR CASCADE classifier to detect facial features. The system was then trained using a Multilayer Perceptron (MLP) classifier on the prepared dataset. This approach enables real-time identification of individuals wearing or not wearing masks. The solution is particularly valuable in enforcing health safety protocols during pandemic conditions.