There are many styles and details of apparel, including coat length, collar design, sleeve length and other detail properties. The e-commerce platform that manages apparel products needs to quickly and effectively identify the attribute categories of apparel for quick retrieval. Apparel image data contains many detailed features that can be easily deformed and occluded. Traditional image recognition technology has been unable to meet the requirements of its classification accuracy. The neural network based on deep learning can classify the fine-grained attributes of complex objects well after training. In this work, we use the apparel image data to train convolutional neural network for the classification of fine-grained attributes. To improve the classification accuracy, we also integrate the results of different models. The experiments show that the results of multi-model fusion are better than those of single model.