基于SE模块改进Xception的动物种类识别

Xception是Inception网络的一种极端化表现，在与Inception v3参数相近的情况下，它能够达到更高的准确度。由于神经网络提取的特征不一定都是有用特征，因此以Xception为基础，将SE(Squeeze and Excitation）模块加入该网络，调整特征通道的权重，使得网络的精确度得到提高。通过实验，融合SE模块的Xception网络训练精确度分别在Oxford-IIIT Pet数据集和CUB_200_2011数据集上提升了1%~1.7%和0.8%~1%，证明了SE模块能够进一步提升Xception的精确度。将改进后的Xception应用到动物种类识别中，根据精确度曲线对实验策略调整改进，最终在测试集上获得95.63%的识别率。

Recognition of Animal Species Based on Improved Xception by SE Module

Xception is an extreme manifestation of the Inception network, which can achieve higher accuracy with similar parameters to the Inception v3. Because the features extracted by the neural network are not necessarily all useful features, based on Xception, SE module is added to the network to adjust the weight of the feature channel, which improves the accuracy of the network. Experiments show that the training accuracy of Xception network fused with SE module increases by 1%~1.7% and 0.8%~1% on Oxford-IIIT Pet dataset and the CUB_200_2011 dataset respectively, which proves that SE module can further improve the accuracy of Xception. The improved Xception is applied to animal species recognition, and the experiment strategy is adjusted and improved according to the accuracy curve. Finally, 95.63% recognition rate is obtained on the test set.