File and Datsets structure overview

File structure

The saved file have fillowing structure:

• dataset_name - The data group calls 'dataset' everywhere in this Wiki
  • 'Images' dataset - the list of images
  • 'Marks' dataset - the list of marks lists that corresponds to 'Images'

For image has been wrote down in 'Images' dataset by number n a list of marks will be write down in 'Marks' dataset upon the same number.

The one file may have any number of data groups (dataset is everywhere in this Wiki)

Images structure

The image is being saved as array with shape (3, img_h, img_w) where:

img_w is image width specified in img_w tag in config

img_h is image hight specified in img_h tag in config

3 is number of color channels. Red, Green, Blue

This array contains color intensity of each channel in each pixel as float value. The intensity values interpolates between 0 and 1 values.

Marks list structure

Coordinates of marks will be write down as array with shape (marks_counts, 2) where:

marks_count the count of marks specified in marks_counts tag of config

2 is coordinates list. First value is x coordinate, second value is y coordinate

Coordinates value have float datatype with zeros afteter comma anyway. So it is actually integer with explict indication to there is no fractional part