Construction of Basketball Teaching Evaluation Model Based on Deep Convolutional Neural Network

Abstract

With the development of network technology, the evaluation methods of basketball teaching and training are constantly innovating. In this paper, a U-shaped encoder-decoder architecture network is adopted. DeepGlobe is used to extract data sets to test the performance of the extraction model constructed by deep convolution neural network. In this study, residual block, dense extended block, Dice loss function, and multi-scale Dice loss function are tested. The results show that the experimental group has achieved remarkable results. Through formative evaluation, students' mastery of technical and theoretical knowledge can be measured, and the teaching process can be adjusted according to the feedback information. Therefore, formative evaluation can promote the mastery of basketball technical movements and theoretical knowledge.