Changyong Zhu

In order to protect intangible cultural heritage and promote outstanding cultural works, this article introduces innovative research on Shen Embroidery using convolutional neural networks. The dataset of Shen Embroidery was preprocessed to augment the data required for experimentation. Moreover, the approach of transfer learning was introduced to fine-tune the recognition network. Specifically, Spatial Pyramid Pooling (SPP) is employed by replacing the avg pool in the MobileNet V1 network, achieving the fusion of local and global features. The experimental results showed that the improved MobileNet V1 achieved a recognition accuracy of 98.45%, which was 2.3% higher than the baseline MobileNet V1 network. The experiments demonstrated that the improved convolutional neural network can efficiently recognize Shen Embroidery and provide technical support for the intelligent development of intangible cultural heritage.

This paper proposes an innovative research approach using convolutional neural networks to address the issues of limited Shen Embroidery works, insufficient image datasets, and low recognition rates. Firstly, the experimental dataset was expanded through methods such as flipping, rotating, and color variations. Secondly, the MobileNet V1 network, based on deep learning, was selected for image recognition of Shen Embroidery. Then, transfer learning was applied to fine-tune the MobileNet V1 network. Finally, the avg pool in the MobileNet V1 was replaced with Spatial Pyramid Pooling (SPP) to achieve the fusion of local and global features. The experimental results showed that the improved MobileNet V1 achieved a recognition accuracy of 98.45%, which was 2.3% higher than the baseline MobileNet V1 network. The experiments demonstrated that the improved convolutional neural network can efficiently recognize Shen Embroidery and provide technical support for the intelligent development of intangible cultural heritage.