Commit 5f72a5b1 authored by Aldo Von Wangenheim's avatar Aldo Von Wangenheim
Browse files

Update README.md

parent 1970b198
......@@ -9,3 +9,29 @@ The predictions from the Appsthetics model are highly correlated with human rati
These promising results indicate that our method can serve as an effective and efficient means for providing objective aesthetics evaluation during the design process.
**Keywords**: Aesthetics, Mobile application, Android, Deep learning, Visual design
# Appsthetics
Appsthetics is a neural network model developed with the use of the fast.ai library to predict the visual aesthetic rating of mobile Android applications' user interfaces.
The present Jupyter Notebook contains all of the code, training and the results achieved by the model and is structured as such:
## Part 1 - Initializations
This section takes care of all the initializations needed, from checking if the current machine's specs are adequate to setting up your Google Drive as a virtual disk so the Notebook can read from and write to it.
Library imports and all the prep work for the dataset also takes place in here.
## Part 2 - ResNet34
This was the starting point of Appsthetics. We load up the ResNet34 model and perform trainings, fine-tunings and evaluate the results for the ResNet34 in this section.
## Part 3 - Less powerful networks...
This section covers the training, fine-tunings and results of the ResNet18 model.
## Part 4 - More powerful networks...
This section covers the training, fine-tunings and results of the ResNet50 and ResNet101 models.
## Part 5 - Saved models
This is a section with notes of the best achieved results for each model trained. For quick future reference.
## Part 6 - Misc.
This section contains some work done for the Bland & Altman test and other things.
Markdown is supported
0% or .
You are about to add 0 people to the discussion. Proceed with caution.
Finish editing this message first!
Please register or to comment