Data

The training sets contain 19 samples and the testing sets contain 9 samples. Each sample includes a variable number of time-varying TIFF images, and a json file that contains positions of the neurons which is the unique numbers. Each file is a 512*512 TIFF image.

Each folder contains a variable number of images. For example, sample 00.00 contains 3,024 images, while sample 00.01 contains 3,048. The image files themselves are numbered, e.g. image00000.tiff, but all the images in a single folder represent the same sample, just taken at different times with different calcium levels. The training labels exist at the sample level, so you’ll use all the images in a single folder to learn the locations of the neurons.

The image on the top represents the time-varying images in the training and testing data. The image on the bottom is the extract "regions of interest" (the pink regions on the right) that correspond to individual neurons, which is the goal to draw circles around the regions that contain neurons.

Reference:
http://neurofinder.codeneuro.org/