Abstract:

The incorporation of live examples of different tasks within a computer vision (CV) course offers multifaceted learning opportunities. This study delves into the educational advantages derived from such an exercise. By engaging in the development of different computer vision systems, students gain practical experience in applying fundamental computer vision principles to real-world scenarios. Through this hands-on approach, students comprehend the complexities of image processing, object detection, and neural network implementation.

As a practical example, a laboratory exercise in a CV course for the face mask detection is described, which fosters a deeper understanding of various face detection algorithms, neural network architectures, and classification techniques, such as ResNet for face detection and MobileNetV2 for face mask classification. Students explore the nuances of pre-trained models, model deployment on resource-constrained devices like Raspberry Pi, and the evaluation of classification performance using key metrics like precision, recall, F1 score, and accuracy. Furthermore, this exercise promotes critical thinking and problem-solving skills as students tackle challenges related to data preprocessing, model optimization, and algorithm selection.

Finally, the analysis focuses on the average student grades, number of students enrolled and successfully completed the CV course, and the number of defended Master of Science (M.Sc.) theses across three academic years: 2020/2021 (without live laboratory exercises), 2021/2022 (without live laboratory exercises), and 2022/2023 (including live laboratory exercises). While the average grades exhibit a slight decrease in the most recent academic year, this can be attributed to the higher enrollment compared to preceding years. Additionally, there's a noticeable rise in the number of defended M.Sc. theses, indicating a growing interest in the CV course.

Keywords:

Deep neural networks, TensorFlow, Resnet, MobileNet, Raspberry Pi.