Best wrapper library for TensorFlow and Theano
January 18, 2019

Best wrapper library for TensorFlow and Theano

Rounak Jangir | TrustRadius Reviewer
Score 8 out of 10
Vetted Review
Verified User

Overall Satisfaction with Keras

Keras is not used across the whole organization, but it is being used by some of our departments only. And in those departments, most of them are using it to do some kind of machine learning task, which basically includes designing and implementing the neural network. I have used this for lots of reasons. All of them were of machine learning fields, like image processing, basic classification, and much more.
  • Performs well when you are doing some implementation which requires neural network implementation and some other deep learning models
  • It has lots of inbuilt tools which you can have clean your data before processing
  • It supports TensorFlow as its backend, so it can easily use GPU
  • As it is a kind of wrapper library it won't allow you to modify everything of its backend
  • Unlike other deep learning libraries, it lacks a pre-defined trained model to use
  • Errors thrown are not always very useful for debugging. Sometimes it is difficult to know the root cause just with the logs
  • It helped me in learning the basic concept of deep learning by having hands-on experience.
  • It has helped us to implement our NN with very little time.
  • It doesn't give you the whole power to customize your neural network. If you want that then you have to shift to TensorFlow
Keras is a good point where you can learn lots of things and also have hands-on experience. There is not much comparison of Keras with Tensorlow, as Keras is a wrapper library which supports TensorFlow and Theano as backends for computation. But once you have enough knowledge of deep learning and machine learning, then it's better if you use TensorFlow itself or some other library.
There are a lot of other libraries in competition with Keras, like TensorFlow, tfLearn, Theano and lot more. If you are new to the deep learning field and want to learn things quickly with implementation then I think you should start with Keras. Once you have good enough knowledge about deep learning concepts then you can shift to some other library, most probably to TensorFlow which gives you the power to write and customize whole neural network