Brett has large set of data with twenty million data points that hold the images of four types of planes. These are the 747, 717, A380 and the Concorde. Every class has four, three, one, and two million data points. Brett wants to train a model that can detect these four planes from images. Brett's most recent trained model gives a validation accuracy of twenty six percent. He deems that this is sufficient due to the fact that it is higher than a random draw of twenty five percent. Is this a correct or incorrect practice?