This study introduces a novel convolutional neural network, the WISE Galaxy Classification Network (WGC), for classifying spiral and elliptical galaxies using Wide-field Infrared Survey Explorer (WISE) images. WGC attains an accuracy of 89.03%, surpassing the combined use of K-means or SVM with the Color–Color method in more accurately identifying galaxy morphologies. The enhanced variant, WGC_mag, integrates magnitude parameters with image features, further boosting the accuracy to 89.89%. The research also delves into the criteria for galaxy classification, discovering that WGC primarily categorizes dust-rich images as elliptical galaxies, corresponding to their lower star formation rates, and classifies less dusty images as spiral galaxies. The paper explores the consistency and complementarity of WISE infrared images with SDSS optical images in galaxy morphology classification. The SDSS Galaxy Classification Network (SGC), trained on SDSS images, achieved an accuracy of 94.64%. The accuracy reached 99.30% when predictions from SGC and WGC were consistent. Leveraging the complementarity of features in WISE and SDSS images, a novel variant of a classifier, namely the Multi-band Galaxy Morphology Integrated Classifier, has been developed. This classifier elevates the overall prediction accuracy to 95.39%. Lastly, the versatility of WGC was validated in other data sets. On the HyperLEDA data set, the distinction between elliptical galaxies and Sc, Scd and Sd spiral galaxies was most pronounced, achieving an accuracy of 90%, surpassing the classification results of the Galaxy Zoo 2 labeled WISE data set. This research not only demonstrates the effectiveness of WISE images in galaxy morphology classification but also represents an attempt to integrate multi-band astronomical data to enhance understanding of galaxy structures and evolution.