Abstract: The TumorSeg project focuses on building a deep learning-based system for detecting brain tumors from medical images. It uses a publicly available dataset containing two classes: Non-Tumor (Class 0) and Tumor (Class 1). Each image is annotated at the pixel level, allowing the model to not only detect the presence of a tumor but also accurately identify its exact location and shape within the brain. The project aims to perform both classification and segmentation tasks effectively. For segmentation, advanced models like UNet2 and UNet3 are used, as they are well-suited for precise pixel-wise prediction. For classification, models such as MobileNet and DenseNet are employed to extract meaningful features and improve detection accuracy. By combining these techniques, TumorSeg enhances the reliability of tumor detection, supporting early diagnosis and better treatment planning. This project demonstrates how AI can significantly improve medical imaging by providing fast, accurate, and automated analysis. Keyword: Brain tumor, semantic segmentation, image classification, deep learning, Unet, MobileNet, DenseNet.
Basham et al. (Thu,) studied this question.