To define the data set we will start by choosing or collecting a suitable data set for the person recognition task. This dataset should contain images or videos of people, along with tags that indicate the identity of each person.

1.1 Choose the Data Type:
Decide what type of data you want to use to recognize people. This data can be images, video frames or even long video sequences, depending on the nature of your project. For example, for facial recognition, you can use images of people's faces.

1.2 Data Collection:
Collect or obtain the data from the right sources. This may involve photographing people in different positions or lighting conditions, or collecting images and videos from other sources. Make sure you obtain the appropriate permission and comply with data protection regulations.

1.3 Data Labeling:
To train a person recognition model, we need to label each data instance with the person's identity. This may involve creating files or metadata that associate each image or video with that person's name or unique identifier.

1.4 Division of the Data Set:
After you have collected and labeled the data, divide it into at least three different sets: the training set, the validation set, and the test set. The training set will be used to learn the model, the validation set to tune the hyperparameters, and the test set to evaluate the final performance of the model.

1.5 Data Augmentation (optional):
You can consider data augmentation to increase the quantity and diversity of your training data. Augmentation involves applying simple transformations to images, such as rotations, scaling, mirrors, or brightness adjustment, to obtain multiple training examples from the same image.

1.6 Organization of the Director Structure:
Organize your data into a well-defined directory structure so you can easily access and keep track of each dataset. For example, you could have a main directory with subdirectories for the training set, validation set, and test set, each with subdirectories for each class or person.