tf-keras-vis 0.1.0

Neural network visualization toolkit for tf.keras

tf-keras-vis

tf-keras-vis is a visualization toolkit for debugging Keras models with Tensorflow 2.0, but not original Keras.

The features of tf-keras-vis are based on keras-vis, but tf-keras-vis's APIs doesn't have compatibility with keras-vis's because, instead of getting it, we prioritized to get following features.

• Support processing multipul images at a time as a batch
• Support tf.keras.Model that has multipul inputs (and, of course, multipul outpus too)
• Allow use optimizers that embeded in tf.keras
• Allow more clear, sharp and stable visualizing in ActivationMaximization

And then we will add some algorisms such as below.

• SmoothGrad: removing noise by adding noise (DONE)
• Deep Dream
• Style transfer

Requirements

• Python 3.6+
• (tensorflow or tensorflow-gpu) >= 2.0

Installation

• PyPI

$ pip install tf-keras-vis

• Sources

$ cd tf-keras-vis
$ pip install -e .

Or

$ cd tf-keras-vis
$ python setup.py install

Usage

T.B.D.

For now, Please see examples/activation_maximization.ipynb and examples/saliency.ipynb. When you want to run jupyter notebook, we recommend that install tf-keras-vis such as follow:

$ cd tf-keras-vis
$ pip install -e .[examples]

API Documentation

T.B.D

Known Issues

• With InceptionV3 ActivationMaximization doesn't work well, that's, it might generate meanninglessly bulr image.
• With cascading model gradcam doesn't work well, that's, it might occur some error.