Concrete crack detection is crucial for maintaining structural integrity and ensuring safety in civil engineering. This study presents a comprehensive approach to detecting concrete cracks using three advanced machine learning models: support vector machine (SVM), convolutional neural network (CNN), and transfer learning with Inception. We evaluate the performance of each model on a concrete crack dataset, reporting accuracy metrics to assess their effectiveness. The SVM model achieved an accuracy of 88.86%, demonstrating its proficiency in distinguishing between cracked and non-cracked concrete surfaces. The CNN model, known for its ability to learn spatial hierarchies in images, achieved a slightly higher accuracy of 89.36%. However, the most significant improvement was observed with the Inception model, which leverages transfer learning to enhance feature extraction and classification. This model achieved an impressive accuracy of 94.10%, outperforming both SVM and CNN.