Title: | A Machine learning model development within MLOps practice |
Subject: | Computer science, Robotics, Software engineering |
Level: | Advanced |
Description: |
The goal of the thesis is to develop the machine learning (ML) part of a smart Emergency Brake Control System, which is used in an autonomous vehicle. The intended system is an ML-enabled software system that is able to detect pedestrians from the input images and sends an emergency brake command upon detecting any pedestrian to a Brake system. The ML model development is the target of the thesis which is done through the steps of an MLOps practice. The input images of the detection system are supposed to be received from the cameras, and in this project, they would be imported from open source ML data sets or a simulation environment for autonomous vehicles such as CARLA. The developed model will be connected to a simulation environment and will be used for the simulation-based testing of an Advanced Driver Assistance System (ADAS). No hardware implementation is intended in the project. The ML system could be implemented and built using “YOLO library” or “Detectron2: A PyTorch-based modular object detection library, which is one of the Facebook AI Research (FAIR)’s open-source projects”. Output: A pedestrian detection system that is implemented within the context of MLOps practice. References: https://cloud.google.com/solutions/machine-learning/mlops-continuous-delivery-and-automation-pipelines-in-machine-learning https://github.com/ultralytics/yolov3 https://ai.facebook.com/blog/-detectron2-a-pytorch-based-modular-object-detection-library-/ https://github.com/facebookresearch/detectron2 Supervisor(s): Mahshid Helali Moghadam, Mehrdad Saadatmand Examiner: |
Start date: | |
End date: | |
Prerequisites: | Python |
IDT supervisors: | |
Examiner: | |
Comments: | |
Company contact: | RISE Research Institutes of Sweden Mahshid Helali Moghadam, mahshid.helali.moghadam@ri.se |