Canine cataracts, which are caused by factors such as genetic predisposition and metabolic irregularities, can lead to complete blindness if not diagnosed early. The project simplifies the diagnosis through photos, helping dog owners. Despite the varied clustering techniques explored during the project, the 3-cluster test showed promising results in cataract identification. It is crucial to note that definitive diagnoses should always be performed by specialists. In the material and methods section, a heterogeneous dataset of dog eye images was used, employing the Ultralytics library and the YOLOv8n model for training. The process included cropping eyes, resizing images, and extracting features with a pre-trained VGG16 model. Principal Component Analysis (PCA) reduced dimensionality, and the K-Means clustering algorithm categorized the data. The results show that the eye extraction process achieved 99.7% precision and 97.4% recall. Evaluation through the Elbow Method and Silhouette Analysis identified the three-cluster configuration as the most effective in distinguishing visual patterns within the dataset. The discussion emphasizes the feasibility of employing Deep Learning for cataract detection, though acknowledging the need for further precision refinement. The project has potential as a tool for dog owners, pending clinical application refinement.