Data Preprocessing

This folder contains the code for data preprocessing. The data preprocessing is done in two steps:

1. rename.py: This script renames the files in the dataset to a consistent format. The format is parition_index_class.jpg. The script takes the path to the dataset as input and renames the files in the dataset.
2. move.py: This script moves the files in the dataset to a common partitioned directory. That is every image in the train partition is in the same train folder. The script takes the path to the dataset as input and moves the files to the partitioned directory.

Usage

The current directory structure is as follows:

rafdb/
    train/
        angry/
            train_1_aligned.jpg
            train_2_aligned.jpg
            ...
        disgust/
        ...
        |
    test/
        angry/
            test_1_aligned.jpg
            test_2_aligned.jpg
            ...
        disgust/
        ...
        |
    val/
        angry/
            val_1_aligned.jpg
            val_2_aligned.jpg
            ...
        disgust/
        ...
        |

Run the rename.py script to rename the files in the dataset:

python rename.py

The new directory structure will be as follows:

rafdb/
    train/
        angry/
            train_1_angry.jpg
            train_2_angry.jpg
            ...
        disgust/
        ...
        |
    test/
        angry/
            test_1_angry.jpg
            test_2_angry.jpg
            ...
        disgust/
        ...
        |
    val/
        angry/
            val_1_angry.jpg
            val_2_angry.jpg
            ...
        disgust/
        ...
        |

Run the move.py script to move the files in the dataset to a common partitioned directory:

python move.py

The new directory structure will be as follows:

rafdb/
    train/
        train_1_angry.jpg
        train_2_angry.jpg
        train_1_disgust.jpg
        ...
        |
    test/
        test_1_angry.jpg
        test_2_angry.jpg
        test_1_disgust.jpg
        ...
        |
    val/
        val_1_angry.jpg
        val_2_angry.jpg
        val_1_disgust.jpg
        ...
        |

CSV Generation

Run the data_csv.py script to generate the CSV file for the dataset:

python data_csv.py

This will generate a CSV file for every partition in the dataset with the following format:

image_name class
train_1_angry.jpg 3
train_2_happy.jpg 0
train_3_neutral.jpg 6
...

And the directory structure will be as follows:

rafdb/
    train/
    train_labels.csv
    test/
    test_labels.csv
    val/
    val_labels.csv

The dataset is now ready for training.