Creation of Materials to Teach Data Science via Self-Driving
Name
Artur Kreegipuu
Abstract
Data science projects often contain pitfalls that students might understand theoretically but rarely experience in practice. Such crucial problems are vividly demonstrable on self-driving toy cars. Students can witness how common data science mistakes during the training and deployment process impact the performance of self-driving neural networks. This thesis focused on creating practical study materials to help students understand, detect, and prevent common pitfalls in data science using self-driving toy cars. Practical tasks that involve gathering data, training and deploying self-driving models, highlighting various machine learning pitfalls and limitations of artificial intelligence, were developed for the educational materials. Practical tasks were tested in various different lighting conditions and expected outcomes were filmed. Feedback from two machine learning experts was collected about the study materials.
Graduation Thesis language
English
Graduation Thesis type
Bachelor - Computer Science
Supervisor(s)
Ardi Tampuu
Defence year
2024