Sunspotter - 14 years of SoHO/MDI

Second dataset based on automated detections produced by SMART [1] with data between May 1996 and August 2010 and classifications by volunteers through the zooniverse's Sunspotter project.
Volunteers had to choose the most complex active region of a pair based on a random selection of the least classified images. Unlike the first round [2], all active regions were scaled to have the same size. Attempting therefore to remove the bias produced by the differences on sizes.

The dataset is composed of five files:

• lookup_timesfits.csv: lists the filenames and the date of the data acquisition.
• lookup_properties.csv: lists the properties about the active region observed in each frame to be classified. Some of the properties as measured by SMART [1]
• classifications.csv: lists each classification made by the volunteers.
• rankings.csv: lists the final ranking on complexity.
• users_links.csv: cross identifiers of users between dataset 2 (this one) and dataset 1 [2].

The score provided on the rankings file follows the Elo rating system. However, a new score following other selection mechanism is possible using the data available on the classification file.

Though the user's information has been removed, the classifications keep an index to differentiate classifications made by different users.

Some software to ingest the tables into a sqlite database and to obtain some preliminary results are available on GitHub.

[1]: DOI: 10.1016/j.asr.2010.06.024

[2]: DOI: 10.5281/zenodo.1478966