xnap-project-matcad_grup_3 Repository

This repository contains the source code for a project divided into two main folders: coral-cnn-master and AgePrediction.

coral-cnn-master

The coral-cnn-master folder serves as the starting point for the project. It includes several important subfolders:

• model-code: This directory houses nine executable files, each responsible for training a model over a span of 200 epochs.

• single-image-prediction_w-pretrained-models: Here, you can find files that enable the prediction of the age of a single image using pretrained models.

• experiment-logs: This folder stores the results of the trained models, including predictions, losses, and model parameters.

For more information about the files in the coral-cnn-master folder or to access the download links, please refer to the coral-cnn-master README.

AgePrediction

The AgePrediction folder contains the following structure:

• models directory: This directory consists of the _init_.py and models.py files. They define classes and functions related to the models used in the project. Many of these classes and functions are generalizations of the models found in coral-cnn-master, with minor enhancements.

• utils directory: Here, you will find the _init_.py and utils.py files, which define various functions used in the execution files. Most of these functions originate from the starting point code.

• sortides directory: This folder serves as the destination for files generated during executions. The --outpath parameter should always be specified as a relative path to this folder.

• main.py: This is the main executable file, encompassing the project's overarching ideas. It allows for modifications to the dataset or loss using flag parameters, initialization of a pre-trained model for improvement, and predictions using the test dataset.

Example

The following code trains coral on the CACD dataset:

python main.py --cuda 0 --seed 1
--outpath ./AgePrediction/sortides/predictions-cacd_coral
--dataset cacd
--loss coral
--starting_params 0
--state_dict_path /home/xnmaster/projecte/AgePrediction/sortides/cacd-pretrained/cacd-coral__seed1/best_model.pt

• --cuda <int>: The CUDA device number of the GPU to be used for training (--cuda 0 refers to the 1st GPU).
• --seed <int>: Integer for the random seed; used for training set shuffling and the model weight initialization (note that CUDA convolutions are not fully deterministic).
• --outpath <directory>: Path for saving the training log (training.log) and the parameters of the trained model (model.pt).
• --dataset <str>: Flag that indicates the dataset to use (CACD, AFAD).
• --loss <str>: Flag that indicates the loss function to use (ce, coral, ordinal).
• --starting_params <int>: Integer to indicate whether or not to use the best parameters of a pretrained model.
• --state_dict_path <directory>: Path which contains the pretrained model file (best_model.pt).