Getting Started with Deep Learning

The next era of image processing is here…

DRAGONFLY FREE 30-DAY TRIAL

Refer to the videos below to help you get started with Dragonfly's Deep Learning tool, which lets you advance your image processing results to new levels.

Additional information about Deep Learning is available in our References and Resources section and in the Deep Learning FAQs. A number of cases studies are also available in Case Studies.

Deep Learning for Imaging Scientists

This video provides the background details for a better understanding of Dragonfly's Deep Learning Tool and how to train deep models for denoising, image segmentation, and super-resolution. Topics include the comparison of image processing to linear regression, fitting and applying functions, perceptron and neural networks as functions, as well as model training and optimization.

References and data examples:
Go to Grant Sanderson's YouTube Channel, 3Brown1Blue, for a deeper explanation on neural networks and how they are trained.
Pharmaceutical tablets example from Ma et al. 2020.
Rat neurons example from Eustaquio et al. 2018.


Denoising with Deep Learning

This video introduces denoising with Dragonfly's Deep Learning Tool and includes the following topics:

  • System requirements for Deep Learning.
  • Extracting training data from marked slices.
  • Generating Deep Models for denoising.
  • Selecting the inputs and setting the training parameters.
  • Selecting the inputs and setting the training parameters.
  • Training denoising models and previewing training results.
  • Questions and answers.

Image Segmentation with Deep Learning

This video introduces image segmentation with Dragonfly's Deep Learning Tool and includes the following topics:

  • System requirements for Deep Learning.
  • Labeling data and creating multi-ROIs from regions of interest.
  • Generating models for binary and multi-phase segmentation.
  • Choosing the training inputs and parameters.
  • Selecting the inputs and setting the training parameters.
  • Training deep models for segmentation and previewing model predictions.
  • Segmenting full datasets.
  • Questions and answers.

Super Resolution with Deep Learning

This video introduces super resolution with Deep Learning and includes the following topics:

  • What is super resolution?
  • Super resolution with U-Net, WDSR-A/WDSR-B, and DenseNet.
  • Previewing and applying super resolution models.
  • Questions and answers.

Publications:
Wang et al. Super Resolution Convolutional Neural Network Models for Enhancing Resolution of Rock Micro-CT Images.
Izadi et al. Can Deep Learning Relax Endomicroscopy Hardware Miniaturization Requirements? (arXiv.org, June 21, 2018).


Advanced Topics in Deep Learning

This video discusses advanced topics in Deep Learning and includes the following:

  • Data augmentation.
  • Transfer learning.
  • Multi-channel segmentation.
  • New features, such as the Segmentation Wizard, multi-slice segmentation, and feature detection with YOLO.
  • Instance segmentation (deep watershed, fiber isolation).
  • Questions and answers.

Has Dragonfly Convinced You?

Start enjoying all Dragonfly included features with a free 30-day trial version.