Deep Learning

Why Use MATLAB for Deep Learning?

MATLAB makes deep learning easy and accessible for everyone, even if you’re not an expert. Design and build your own models, access the latest models, or import pretrained models from Caffe. Use NVIDIA GPUs to train your models. Automatically generate high-performance CUDA code for embedded deployment.

Network Architectures and Algorithms

  • Convolutional Neural Networks (CNNs) for image classification, object detection, regression, and semantic segmentation
  • New Directed Acyclic Graphs (DAGs) networks to represent complex architectures
  • New Long short-term memory (LSTM) networks for prediction and classification on time-series, text, and signal data

Training and Visualization

  • New Monitor training progress with plots for accuracy, loss, validation metrics, and more
  • New Author custom layers in MATLAB using the Custom Layer API
  • New Automatically validate network performance, and stop training when the validation metrics stop improving
  • New Perform hyperparameter tuning using Bayesian optimization
  • Visualize activations and filters from intermediate layers
  • Use Deep Dream visualization

Access the Latest Pretrained Models

  • New TensorFlow-Keras model importer; GoogLeNet and ResNet-50 models
  • Import models from Caffe (including Caffe Model Zoo)
  • VGG-16, VGG-19, and AlexNet models
  • Coming soon: InceptionV3 model

Scaling and Acceleration

  • Accelerate training using single and multiple NVIDIA GPUs
  • Scale training to cloud and clusters using Amazon EC2

Handling Large Sets of Images

  • Use imageDatastore to easily read and process large sets of images
  • Access data stored in local files, networked storage, databases, and big data file systems
  • New Efficiently resize and augment image data to increase the size of training datasets

Object Detection

  • Train object detectors using R-CNN, Fast R-CNN, and Faster R-CNN algorithms
  • New Evaluate detector performance including precision and miss-rate metrics

Semantic Segmentation

  • New Classify individual pixels using sematic segmentation
  • New Validate the performance of semantic segmentation algorithms

Ground-Truth Labeling

  • New App to label pixels and regions for semantic segmentation and object detection
  • Automate ground-truth labeling using automation API

Embedded Deployment

  • New Automatically convert deep learning models in MATLAB to CUDA with GPU Coder
  • New Run trained models up to 4.5x faster than Caffe2 and up to 7x faster than TensorFlow

Get a Free Trial

30 days of exploration at your fingertips.

Have Questions?

Talk to a deep learning expert.