Deep Learning

Tools and Libraries

The main libraries for building deep learning models are:

• TensorFlow
• PyTorch
• Caffe
• MXNet
• Neon (Intel framework)
• DeepLearning4J (for Java developers)

If you aren’t a programmer there are also higher-level abstractions to help interact with these libraries. Here are some of the most common ones built on Theano.

• Lasagne: https://lasagne.readthedocs.io/en/latest/
• Keras: https://keras.io/
• Blocks: https://blocks.readthedocs.io/en/latest/

Review

Wikipedia page comparing libraries: https://en.wikipedia.org/wiki/Comparison_of_deep_learning_software

Activity on Github for all the major libraries:

Comparison table

(Comparative Study of Deep Learning Software Frameworks)

Tutorials

web book

• A Guide to Deep Learning (general): link
• A Beginners Guide to Understanding CNNs: link

Pretrained Networks

Network Type	Dataset	Problem Domain	URL
U-Net	Lung, CT	Segment lung nodules	https://github.com/booz-allen-hamilton/DSB3Tutorial
U-Net	Electron microscopy	Cell segmentation	http://lmb.informatik.uni-freiburg.de/people/ronneber/u-net/
VGG16, Inception, Xception, ResNet…	ImageNet	Classifying real-world images	https://keras.io/applications/