To identify viruses in transmission electron microscopy (TEM) pictures, this research introduces a new two-pipeline method that leverages convolutional neural networks (CNNs). We used a collection of viral photos to train two separate convolutional models, improving feature extraction using a variety of image processing methods. A Quadratic Weighted Kappa (QWK) score of 0.9719 and a maximum testing accuracy of 97.44% indicate that the suggested model is very reliable for viral classification tasks. An average F1-score improvement of 5.32% and a substantial improvement of 5.24% in accuracy compared to conventional single-pipeline approaches were produced by the dual pipeline approach. In addition, confusion matrices showed that the model performed well for most categories of viruses, even if certain classes were difficult to accurately classify. Consistent with and expanding upon previous research, our results demonstrate that deep learning methods are capable of accurately differentiating between different types of viruses. This study demonstrates the promise of sophisticated convolutional neural network (CNN) designs in the field of virology and sets the stage for further research into how to improve the efficiency and accuracy of viral classification for better disease management and response.