In the field of education, the application of Big Data (BD) has become more and more extensive. These data can not only help educators better understand students’ learning characteristics and problems, but also provide important support for personalized teaching and precise management. The purpose of this paper is to study a 3D reconstruction algorithm of environmental design image based on Convolutional Neural Network (CNN), and compare its performance with that of Support Vector Machine (SVM) in modeling accuracy and user experience. By constructing a multi-level CNN structure and automatically learning and extracting features from the original image, this algorithm can effectively suppress background information and enhance feature information related to environmental design. The experimental results show that the modeling accuracy of this algorithm is improved by 24.82%, which shows that this algorithm can better restore the feature information of the environment image and suppress the background information. Moreover, the instructional platform constructed by the algorithm in this paper has a high score among learners, which shows that the instructional platform has high practical value and popularization value, and can provide strong support for education and training in the field of environmental design.