The segmentation of oral histology images presents a significant challenge in dental research and education due to the scarcity of high-quality labeled datasets and the inherent complexity of multi-class tissue structures. Manual annotation of histological slides is a labor-intensive and expert-driven process, making large-scale data acquisition difficult. To address this limitation, we propose a multi-model deep learning framework for supervised multi-class segmentation of microscopic oral histology images. In particular, we propose in this paper 4 main contributions. First, we construct a custom annotated proprietary dataset of 32 images using Roboflow. Second, we extensively augment the dataset using 6 combined morphological transformations to get a total of {6122 annotated images}. Then, we generate annotations in COCO-format along with gray-level masks to support pixel-level class differentiation across diverse anatomical structures. Third, we train using Tensorflow multiple convolutional neural network (CNN)-based architectures, including U-Net, Mask R-CNN, SAM, DeepLab, MobileNet, Yolov8, and transfer learning from microscopy-specific models for the segmentation task. We evaluate their performance using standard segmentation metrics such as mean Intersection-over-Union (IoU), Dice Coefficient, Pixel Accuracy, and mean Average Precision (mAP). Forth, we develop a Flutter-based mobile application to extend practical usability. This application enables students and faculty to interact with the system through features such as segmentation testing and institutional announcements. Our work establishes a modular, scalable, and user-centric foundation for advancing AI-assisted analysis in dental histology.