Keras Application Zoo

Introducing Keras Application Zoo: A library for reusable deep learning models in Keras.

Keras Application Zoo is a public clearinghouse to publish, discover, and reuse parts of machine learning modules in Keras. By a module, we mean a self-contained piece of a Keras Applications-like model, along with its weights, that can be reused across other, similar tasks. By reusing a module, a developer can train a model using a smaller dataset, improve generalization, or simply speed up training.

Lots of researchers and engineers have made their deep learning models public in various frameworks for different tasks with all kinds of architectures and data. These models are learned and applied for problems ranging from simple regression, to large-scale visual classification.

However, Keras does not contain the degree of pre-trained models that come complete with Caffe.

To lower the friction of sharing these models, we introduce the Keras Application Zoo:

  • A central GitHub repo for sharing popular deep learning models with Keras code & weights files
  • Contains ONLY additional deep learning models which are not yet available within keras.applications module itself or Keras community contributions official extension repository
  • Tools to upload/download model info to/from GitHub, and to download trained Keras Applications-like binaries
  • Models can be used for prediction, feature extraction, and fine-tuning just like the genuine canned keras.applications architectures
  • No separate models configuration files in a declarative format. Models are described in Python code, which is compact, easier to debug, and allows for ease of extensibility

BENEFIT FROM NETWORKS THAT YOU COULD NOT PRACTICALLY TRAIN YOURSELF BY TAKING KERAS TO THE ZOO!

Related