An electronics manufacturing company is building an image classification model to detect whether a circuit board is defective or not. To decrease training time, a Machine Learning Specialist implemented transfer learning using 300 images on a pre-trained neural network obtained from a partner company. The company requires a model accuracy of no less than 95% to be feasible. The Specialist conducted a full grid search through a wide hyperparameter space. However, the optimal values returned only 73% and 72% accuracy on the training and validation set, respectively. How can the Specialist increase the model accuracy?