datasets.zip

Datasets of
confocal microscopy images of cardiomyocytes aimed at development of image
recognition systems. Images of live and healthy cardiomyocytes were assigned
into 5 classes according to their development stages. The least developed
cardiomyocytes were considered to be at stage~1 (class 1) while the most
developed ones were assigned stage~5 (class 5). All other images belong to a
class 0.

We provide
three datasets of the cardiomyocyte images. The resolution of images is 1024 x
1024 px. Datasets T-XXX-A and T-XXX-B are aimed at training of image
recognition systems, while a dataset E-XXX-A is aimed at their evaluation. The
dataset T-XXX-B is T-XXX-A extended about manually modified images, so that T-XXX-B
is 3 times bigger compared to T-XXX-A. We also provide a Python script which
allow augmentation of the images by their flipping and rotation. The script
further allows resizing of the images and it generates a csv file that consists
of records about each generated image. The records specify all operations that
were used for its generation.