Sooraj Abraham

— Optimized LiDAR 3D point cloud object detection network (PointPillar) for embedded devices. — Analyzed model performance with pruning at different sparsities and quantization. — Improved network runtime by 3.72 times, reducing processing time from 1600 milliseconds to less than 430 milliseconds, with a minimal accuracy drop of 4%.

— Prototyped In-Cabin Monitoring System for driver drowsiness and emotions. — Deployed Jetson Nano and optimized software performance for real-time monitoring. — Implemented drowsiness detection and fine-tuned emotion recognition models using TensorFlow and CNN, achieving a 92% accuracy rate in driver emotion classification.