The development of cross-border e-commerce logistics services has injected new vitality into the development of international trade, and therefore has become a new hot spot in theoretical research. In order to ensure the healthy development of cross-border e-commerce, it is urgent to build a set of scientific and effective evaluation mechanisms to scientifically evaluate the logistics service quality of cross-border e-commerce. Multi-angle perceptual convolutional neural network is a framework for service scene identification of cross-border e-commerce logistics enterprises based on deep convolutional neural network and multi-angle perceptual width learning. In this article, both shallow features and deep features were input into the deep perception model (DPM) to obtain a set of distinguishable features with causal structure, which was used to completely describe the high-level semantic information of cross-border e-commerce logistics enterprise services. Among them, DPM mainly adopts the fusion strategy of shallow feature and deep feature. Meanwhile, the feature representation is input into the width learning pattern recognition system for training and classification, so as to evaluate the service quality of cross-border e-commerce logistics enterprises. The multi-angle perceptual convolutional neural network can effectively solve the problems of high similarity between service classes of cross-border e-commerce logistics enterprises and large differences within the class, and achieve better generalization performance and algorithm complexity than support vector machine, random forest and convolutional neural network.