In R2017a, MATLAB® makes it easier and faster for engineers and scientists to learn and apply deep learning to computer vision problems. Configure and train models, visualize their structure, leverage pretrained models for transfer learning, and take advantage of GPU acceleration.
You can train deep learning models by using a large set of labeled data and deep neural network architectures that contain many layers. Visualization tools help you to understand the structure of the model and see the features learned.
Pre-trained models can be used for transfer learning, which is the process of fine-tuning a model to work on a new data set.
GPUs are highly efficient on parallel algorithms such as deep learning. You can achieve higher levels of parallelism by using multiple GPUs or by using GPUs and processors together.
To handle large sets of images, you can create an
ImageDatastore, which allows you to read and process multiple image files as a single entity.