Thesis Proposal -Real- Time Vision-Based Passenger Monitoring on Edge AI Hardware for Autonomous Buses

Autonomous buses require onboard computing capable of processing sensor data and responding to events in real time. Vision-based

passenger monitoring can require significant computational resources, particularly when multiple camera streams and deep learning

models are used.

Edge AI platforms such as the NVIDIA Jetson Nano enable computer vision and machine learning to run directly onboard the vehicle,

reducing latency and network requirements. However, limited computing, memory, and power resources create challenges for real-time

deployment.

This thesis aims to develop and evaluate an efficient vision-based passenger monitoring pipeline for edge AI hardware. The work

will investigate how computer vision and deep learning models can be optimized to achieve a suitable balance between detection

accuracy and computational efficiency.

• Investigate model optimization and hardware acceleration using techniques such as quantization, reduced resolution, CUDA, and

TensorRT.

• Evaluate accuracy, latency, resource usage, and scalability across different camera configurations and realistic autonomous-bus

scenarios.

The work will include model selection, implementation, optimization, and evaluation, with a focus on balancing monitoring

performance and real-time computational requirements.

Meritorious: Experience with NVIDIA Jetson, CUDA, TensorRT, PyTorch, TensorFlow, OpenCV, ROS2, model optimization, embedded

systems, or real-time computer vision.

Tags: Autonomous Vehicles, Autonomous Bus, Edge AI, NVIDIA Jetson, Computer Vision, Deep Learning, Real-Time AI, Model

Optimization, Passenger Monitoring

To give you the best possible support during your thesis, we’d like you to be able to come to the office connected to the project

and spend most of your time working from there.

We look forward to receiving your resume, and preferably, a personal letter in which you explain why you want to write your thesis

with Syntronic.