Snakes are carnivorous reptiles specifically found in forest areas. There are around 3000 snake species, and they are generally categorized as either venomous or non-venomous. There are some significant challenges for identifying each of them based on their visual features and is a much more complex task for normal humans or even for expert herpetologists as well. In continuation to the prior research based on the visual features of snakes, this paper presents a more accurate and more comprehensive alternative to the ML-based approach. In this study, a fusion of CNN layer and Capsule Network are introduced for snake-image classification. The proposed method uses primary capsules to group the outputs of convolution filters. For experimentation, Indian snake data of around 2000 snake images are used. For comparison, image features are also fed into traditional classifiers such as Support Vector Machines (SVM), Decision Trees (DT), and k-Nearest Neighbors (KNN). The evaluation results indicate that the SVM classifier achieves the highest accuracy, reaching approximately 90.8%, outperforming both DT and KNN classifiers.