High intra-class variance and inter-class similarity are common problems for traditional image classification models, particularly when it comes to differentiating between closely related animal species. By creating a strong classification model intended to classify three visually similar animal classes—dogs, wolves, and foxes—this study tackles these issues. The model uses an enhanced ResNet-18 architecture supplement to control intra-class differences and inter-class similarities. complemented with blocks for spatial attention and squeeze-and-excitation (SE). SE blocks highlight crucial channel-wise information to enhance feature extraction and reduce confusion between related breeds and species. While Spatial Attention blocks focus on significant spatial areas. We have created our own dataset, which contains images of five breeds for each species. The dataset is further divided into training, validation, and testing sets. The developed model has the ability to classify animal classes effectively, even though there are similarities between different classes and variations among animals within the same class.