Ing. Mohamed Elbassiouny

•Implemented custom UNET for jaw segmentation by providing it with the Maximum Intensity Projection(MIP) along the sagittal axis of the DICOM volume. •Utilized YOLO to detect and label teeth of DICOM 2D slices, and perform Z-axis interpolation on 2D slices bounding boxes to generate 3D bounding boxes. •Deployed these models in C++ using ONNX runtime, with custom-built C++ functions for post and pre-processing of data.

• Designed an innovative tracking approach utilizing motion prediction deep learning models based on transformers, integrating interactions among vehicles and between vehicles and the road, instead of traditional Kalman filters. • Deployed a 3D segmentation model. • Developed deep learning models and classical methods to extract lane lines from road using lidar intensity and geometric futures of the road.