Home

Awesome

Explanation_by_Progressive_Exaggeration

Official Tensorflow implementation of ICLR 2020 paper: Explanation By Progressive Exaggeration.

Paper

<img src="./imgs/Model.png" height="50%" width="50%" >

Experimental Results

Qualitative Results

<img src="./imgs/Quality.jpg" height="50%" width="50%" >

Age-Attribute Experiment

Training graphs and qualitative results

./notebooks/Process_Classifier_Output.ipynb

Installation

$ pip install -r requirements.txt

Usage

  1. Download the CelebA dataset and create train and test fold for the classifier.

CelebA dataset is downloaded and saved at ./data/CelebA. IPython notebook creates text files with file names and labels and save them at ./data/CelebA/. These text files are used as input data to train the classifier.

./notebooks/Data_Processing.ipynb
  1. Train a classifier. Skip this step if you have a pretrained classifier. The output of the classifier is saved at: $log_dir$/$name$.

2.a. To train a multi-label classifier on all 40 attributes

python train_classifier.py --config 'configs/celebA_DenseNet_Classifier.yaml'

2.b. To train a binary classifier on 1 attribute

python train_classifier.py --config 'configs/celebA_Smile_DenseNet_Classifier.yaml'
  1. Process the output of the classifier and create input for Explanation model by discretizing the posterior probability. The input data for the Explanation model is saved at: $log_dir$/$name$/explainer_input/
./notebooks/Process_Classifier_Output.ipynb
  1. Trainer explainer model. The output is saved at: $log_dir$/$name$.
python train_explainer.py --config 'configs/celebA_Young_Explainer.yaml'
  1. Explore the trained Explanation model and see qualitative results.
./notebooks/Test_Explainer.ipynb
  1. Save results of the trained Explanation model for quantitative experiments.
python test_explainer.py --config 'configs/celebA_Young_Explainer.yaml'
  1. Use the saved results to perform experiments as shown in paper
./notebooks/Experiment_CelebA.ipynb

Cite

@inproceedings{
Singla2020Explanation,
title={Explanation  by Progressive  Exaggeration},
author={Sumedha Singla and Brian Pollack and Junxiang Chen and Kayhan Batmanghelich},
booktitle={International Conference on Learning Representations},
year={2020},
url={https://openreview.net/forum?id=H1xFWgrFPS}
}