Explaining Generative Diffusion Models via Visual Analysis for Interpretable Decision-Making Process

Code Files

DF-CAM_auc.py: Computes the AUC score for DF-CAM.
DF_CAM.ipynb: Visualizes diffusion models using DF-CAM (Figure 8).
conda_requirements.yaml: If you use Conda, utilize this file for environment setup.
DF_Crossattn.ipynb: Computes cross-attention scores (Figure 9, Figure 11).

DF_LIME.ipynb: Implements the Lime method for diffusion models as a baseline. (Figure3)

Experiments.ipynb: Contains DF-RISE visualization experiments and exponential scheduler experiments (Figure 6, Figure 7, Figure 8, Figure 10).

requirements.txt: Contains the list of packages required for environment setup.
DF_RISE_auc.py: Computes the AUC score for DF-RISE.

Clone this repository:

git clone https://github.com/ian-jihoonpark/X-Diffusion.git

   pip install -r requirements.txt

Name		Name	Last commit message	Last commit date
Latest commit History 23 Commits
figures		figures
models		models
DF-CAM_auc.py		DF-CAM_auc.py
DF_CAM.ipynb		DF_CAM.ipynb
DF_Crossattn.ipynb		DF_Crossattn.ipynb
DF_LIME.ipynb		DF_LIME.ipynb
DF_RISE_auc.py		DF_RISE_auc.py
Experiments.ipynb		Experiments.ipynb
README.md		README.md
conda_requirements.yml		conda_requirements.yml
requirements.txt		requirements.txt
ssim.py		ssim.py
utils.py		utils.py