Config.py

import torch

"""
Define the software configuration. 
"""


class Config:
    """
    Images will be transformed into this shape (pixel height, pixel width) before being fed to the model.
    We recommend (80, 120). More pixels can allow the model to become more precise but it will require more data.
    """

    input_image_shape_basic_model = (80, 160)

    """
    The pretrained model can leverage more parameters, since it doesn't require as much data as training a model from scratch.
    """
    input_image_shape_pretrained_model = (126, 256)

    """
    The folder that contains ant images without background in .png.
    """
    images_folder_path = "clean"

    """
    The original images (with background and scale)
    """
    original_images_folder_path = "original"

    """
    Training is done on a CSV file that contains 5 columns, for example:
    Image Name,x1,y1,x2,y2
    ...
    casent0904793_p_1.png,333,182,610,172
    ...
    (x1, y1) is the where the Vertex of the Head lies and (x2, y2) is where the Petiole lies (pixel coordinates on the image)
    """
    coords_file_path = "CSVs/combined.csv"

    """
    Computer device : mps for mac cuda for windows graphics
    Do not bother if you are using cpu
    """
    device = "mps"
    # we could eventualy just use Config.device everywhere
    is_available = torch.backends.mps.is_available
    #is_available = torch.cuda.is_available # for cuda