Update README.md

README.md

@@ -1,40 +1,7 @@
-# GitHub Repository for XAIDIFF
+# Model Repository for XAIDIFF
 A framework for local example-based explanations with generative models.
 The code is published at [Github](https://github.com/cairo-thws/XAIDIFF)
 The pretrained models can be found at [Huggingface](https://huggingface.co/cairo-thws/XAIDIFF).
 
 ![XAIDIFF Framework](XAIDIFF.png)
 
-
-
-# About this repository
-This repository contains the code for the XAIDIFF framework.
-The framework is designed to provide local example-based explanations for image- and object classifiers with the help of generative models.
-Using a diffusion model, the framework generates counterfactual examples that are similar to the input image but are classified differently by the classifier.
-The framework is designed to be modular and can be used with different generative models and classifiers.
-
-# Installation and usage
-
-## Conda environment
-First, create a new conda environment and activate it.
-```
-conda env create -f environment.yml
-conda activate xaidiff
-```
-
-## Sampling
-Per default the project allows you to sample from any pretrained diffusion model with any pretrained classifier.
-The `SportBallsDataset` is a simple example dataset that can be used to test the framework.
-If you want to add your own classifier and dataset, just inspect the `SportBallsDataset`.
-
-The file `experiment_sampling.py` contains the core functionality of this project and can be used to sample from the models.
-There are many configurable parameters at the head of the file that can be used to adjust the sampling process.
-It will generate counterfactual examples for the given input image and classifier.
-
-The results contain the following information:
-- An image with four rows: the original image, the counterfactual sample, an affirmative sample and an adversarial sample.
-- Different metrics for the generated samples, including LP distances and bits per dimension. This allows to compare the generated samples with the original image.
-
-The file `experiment_sampling_sportsballs.py` contains an example of how to adapt the original sampling process to the `SportBallsDataset`.
-To execute it, download the pretrained sportballs classifier and diffusion model.
-