Research on Innovative Design of Traditional Clothing Handicraft under Computer Big Data Technology

This design implements a traditional clothing handicraft recognition system based on computer big data technology model. Tensor Flow is used to train a custom network, and weight parameters are fixed-point processing. The ARM side OpenCV is used for image preprocessing, and the FPGA side CNNIP is used for real-time recognition. The spatial attention mechanism is introduced into the output feature map of the VGG16 feature extraction network to increase the attention to the clothing area in the image. A bilinear pooling method with a single feature path is designed. While reducing the number of parameters and computation, two operations, global average pooling and global max max pooling, are used to fully my fine-grained features. This system can accurately identify and display pictures of traditional clothing crafts in real time, with an accuracy rate of 92.39%. Under the operating frequency of 100MHz, the image processing speed can reach 1.361ms per frame, and the power consumption is only 0. 53W.