Recently, a new alarming problem has emerged, particularly in residential areas, where dogs and other animals have inflicted substantial harm on children, the elderly, and others arising from stray animals. The severity of such incidents has escalated, leading to serious injuries and even fatalities. In India, these incidents have occurred within the confined environment of residential lifts. To address this pressing issue, this paper proposes an innovative solution: implementing an image classification-based identification system. In this study, ten distinct animal classes are classified using the Animals-10 dataset. To implement this idea, pre-trained architectures like ResNet50 and EfficientNetB7 have been used. Improved results were achieved by fine-tuning these models. The metric used for evaluation was accuracy, ensuring the highest level of improvement. This work achieved an accuracy of 93.53% using ResNet50. However, we achieved the best results using the EfficientNetB7 model and fine-tuning it further. The maximum test accuracy obtained is 98.52%. A comparative study with the previous work is included to demonstrate the effectiveness of the work.