An Improved Method of Image Recognition with Deep Learning Combined with Attention Mechanism

An improved convolutional neural network (CNN) recognition model is proposed for the problems involving low recognition rate and weak generalization ability for flower images. Highly abstracted features after multiple convolutions are integrated, and the performance of network is improved by adding the network model for multi-attention mechanism after residual module for Inception-resnet-V2 Network and fully connected layer before activating the function. The improved model is simulated by integrating OxFlowers 17 and Oxford 102 flower data sets. The results show that the recognition rate of the model based on Inception-resnet-V2 Network combined with attention mechanism is up to 97.6%, being 5.1% higher than that of the original model, and the accuracy for flowers recognition is improved significantly.