Improving CNN Model for Residential Building Image Classification: Enhancing Parameter Estimation Accuracy Through Transfer Learning and Reducing Model Complexity with MobileNet

With the widespread application of deep learning and Convolutional Neural Networks (CNN) in image classification, how to effectively improve model performance and reduce its complexity has become a significant research direction. This study presents an improved CNN model for residential building image classification, which incorporates transfer learning and the MobileNet architecture. Through transfer learning, we effectively enhance parameter estimation accuracy, and by adopting the MobileNet architecture, we reduce the computational complexity and size of the model. Experimental results demonstrate that this approach achieves superior performance on the Cultural Heritage Apps dataset, verifying its efficacy in residential building image classification.