Advanced Machine Learning Project - Domain-To-Text

Dataset

The PACS dataset could be reached from the Datasets google drive folder. It is saved in a .zip format and in the notebook there is no need to to save it on your google drive environment. In any case you can of course download it: https://drive.google.com/drive/folders/1pQlXKQSBjcBPqHQHxTl1bfgs9TRr2tk3?usp=sharing

Pretrained models

In order to reproduce the values reported in the table (see section 1 of the project specifications), you have to use the initial outputs which could be reached from the outputs. They are saved in a .zip format and in the notebook there is no need to save them on your google drive environment. In any case you can of course donwload them: https://drive.google.com/drive/folders/1mjMaP4loQtK5DB26tlv3uvn4rXmDbmk4?usp=sharing

Environment

1. All the requirements for the project are listed under the file requirements.txt;
2. In the folder describing_textures there is the model used to get the new weights;
3. The notebook associated to the project is: https://colab.research.google.com/drive/1LyDXBj6gS-bt937Rzcgc6YWQ8zsviiTS#scrollTo=gi1t1dKcNeS2

Report

For a complete description of the project download the report.pdf from this repository.

Students

The project was developed by:

• Stefano Rainò, s282436
• Alberto Castrignanò, s281689
• Michele Masciavè, s279110
• Lorenzo Mercurio, s290903