tf-keras-vis 0.6.2

Neural network visualization toolkit for tf.keras

tf-keras-vis

tf-keras-vis is a visualization toolkit for debugging tf.keras models in Tensorflow2.0+. Currently supported methods for visualization include:

• Activation Maximization
• Class Activation Maps
  • GradCAM
  • GradCAM++
  • ScoreCAM
  • Faster-ScoreCAM
• Saliency Maps
  • Vanilla Saliency
  • SmoothGrad

tf-keras-vis is designed to be light-weight, flexible and ease of use. All visualizations have the features as follows:

• Support N-dim image inputs, that's, not only support pictures but also such as 3D images.
• Support batchwise processing, so, be able to efficiently process multiple input images.
• Support the model that have either multiple inputs or multiple outputs, or both.
• Support the mixed-precision model. *

And in Activation maximization,

• Support Optimizers that are built to tf.keras.

Visualizations

Visualizing Dense Layer

Visualizing Convolutional Filer

GradCAM

The images above are generated by GradCAM++.

Saliency Map

The images above are generated by SmoothGrad.

Requirements

• Python 3.6-3.9
• tensorflow>=2.0.2

Installation

• PyPI

$ pip install tf-keras-vis tensorflow

• Docker (container that run Jupyter Notebook)

$ cd tf-keras-vis
$ docker build -t <TAG> -f dockerfiles/gpu.Dockerfile .
$ docker run --gpus all --privileged -itd -p 8888:8888 <TAG>

Or

$ docker run --gpus all --privileged -itd -p 8888:8888 keisen/tf-keras-vis:0.5.0-gpu

You can find other images at Docker Hub.

Usage

Please see below for details:

Getting Started Guides

• Saliency and CAMs
• Visualize Dense Layer
• Visualize Convolutional Filer

[NOTE] If you have ever used keras-vis, you may feel that tf-keras-vis is similar with keras-vis. Actually tf-keras-vis derived from keras-vis, and both provided visualization methods are almost the same. But please note that tf-keras-vis APIs does NOT have compatibility with keras-vis.

Use Cases

• chitra
  • A Deep Learning Computer Vision library for easy data loading, model building and model interpretation with GradCAM/GradCAM++.

Known Issues

• With InceptionV3, ActivationMaximization doesn't work well, that's, it might generate meaninglessly blur image.
• With cascading model, Gradcam and Gradcam++ don't work well, that's, it might occur some error. So we recommend, in this case, to use FasterScoreCAM.
• channels-first models and data is unsupported.
• With a mixed-precision model that has a layer which are set float32 dtype exlicitly, ActivationMaximization may raise a error.
• With a mixed-precision model, Regurarization values that is calculated by ActivationMaximization may be NaN.

ToDo

• Guides
  • Visualizing multiple attention or activation images at once utilizing batch-system of model
  • Define various score functions
  • Visualizing attentions with multiple inputs models
  • Visualizing attentions with multiple outputs models
  • Advanced score functions
  • Tuning Activation Maximization
  • Visualizing attentions for N-dim image inputs
• API documentations
• We're going to add some methods such as below.
  • Deep Dream
  • Style transfer