In the past few years, deep learning based methods has quickly become the state of the art in image classification and object detection. As one of the best deep learning structures, Convolutional Neural Network (CNN) is highly automated and requires little prior knowledge. Also, a customized CNN can be quickly built without a large database, if a pre-trained network is provided. These advantages make CNN suitable for undergraduate research. Funded by an 1890 Land Grant Research Project III, CNN is introduced to the undergraduate students in our institution and the students are trained to develop customized CNN in order to solve given image classification problems. The achieved goals and discovered issues are reported and discussed in this work. Overall, the results demonstrated a positive example of integrating modern technology and research into undergraduate classrooms.
展开▼