Facial acne is a prevalent dermatological condition with significant physical and emotional repercussions, especially among adolescents. To address the clinical need for accurate and efficient diagnosis, this study proposes a deep learning-based computer vision model for automatic acne detection in facial images. The methodology consisted of four main phases: Data acquisition (using the public dataset Acne New Data); Preprocessing (resizing, normalization, CLAHE, and standardization); Model implementation (training MobileNetV2, Vision Transformer, ResNet-50, EfficientNetV2-Small); and Evaluation (using Accuracy, Precision, Recall, and F1-Score). The models were trained under uniform experimental conditions using stratified cross-validation. The best result was achieved by the EfficientNetV2-Small model, which obtained 95.43% Accuracy, 95.55% Precision, 92.31% Recall, and a 93.90% F1-Score, demonstrating its robustness and suitability for clinical deployment. These findings confirm the potential of lightweight deep learning architectures in the accurate and efficient classification of facial acne.