Update README.md

README.md

@@ -1,26 +1,51 @@
 ---
 license: mit
 ---
+# Model Card: [InfernoSaber]
+
+## Overview
+
 Welcome to the InfernoSaber, an Automapper for BeatSaber with fully adjusteable difficulty.
 
+### Model List
+
 Recommendation of the models/branches in the following order:
 - expert_15: Current favorite, trained on curated high difficulty maps (8+ nps) with a like/dislike rate >90%. Usually generates good flow.
 - easy_15: Trained on curated low difficulty maps (5- nps) with a like/dislike rate >90%. More creativity but less flow, especially for high difficulty maps.
-- pp3_15: Trained on random maps from my personal collection.
+- pp3_15: Trained on random maps from my personal collection, mainly ranked high PP maps.
+
+### Model Details
 
-The models and corresponding source code are *free to use*, selling the generated maps is prohibited (also labeling as human made is against the beatsaver policies):
-https://github.com/fred-brenner/InfernoSaber---BeatSaber-Automapper
+- **Model Name**: InfernoSaber
+- **Model Version**: v.1.7.0
+- **Architecture**: Multiple custom DNN and Autoencoder with TF v.15
+- **Training Objective**: Classification/Creation of song maps
+- **Language**: Python
+- **License**: Model and source code are free to use
+- **Repository**: https://github.com/fred-brenner/InfernoSaber---BeatSaber-Automapper
 
-This repository contains the trained models for inference (in the branches apart from main).
+## Intended Use
+
+This repository contains the trained models for inference (go to the other branches at the top: "Files and versions", click on "main" -> select "expert_15" or similar).
 All models are trained on different datasets with mixed genre.
+The models are required by the source code at github.
 
+## Inference and Training
 
-Inference can be run on most systems and does not require GPU.
+The models can be run on most systems and do not require GPU for inference.
 
-*Only for training* the following specs are required:
+For training / creation of new models, the following specs are required:
 - 10-20 GB free RAM per 50 maps (mainly depending on the creativity of the maps)
 - 8-15 GB VRAM per 50 maps (mainly depending on the creativity of the maps)
 
+## Limitations and Ethical Considerations
+
+The models and corresponding source code are *free to use*.
+Selling the generated maps is prohibited and labeling as human made is against the beatsaver policies.
+
+This project is not meant to be replacing human maps, but as extension to cover more unkown maps and massively reduce the amount of human effort required.
+
+## Contact Information
 
 If you have comments, suggestions, or want to contribute to new models/features feel free to open a discussion.
-Enjoy!
+Enjoy!