Cancer is one of the most devastating health conditions worldwide, resulting from uncontrollable cell divisions that invade adjacent tissues and organs. Early detection is crucial to improving patients’ quality of life and, sometimes, enabling a cure. In this context, computer vision tools hold significant potential to expedite cancer diagnosis. This study proposes analyzing a double transfer learning approach using histopathological images, where the model is initially trained on ImageNet, followed by fine-tuning with BreaKHis pre-trained weights before applying it to veterinary images. The results indicate that double transfer learning improves model performance by approximately 5%, as observed in k-fold cross-validation with k = 5, highlighting the effectiveness of this approach in enhancing the accuracy and robustness of models for cancer diagnosis in veterinary imaging.