Disk Detective was launched in January 2014, and was expected to continue till 2017. In April 2019 Disk Detective uploaded partly classified subjects, as Zooniverse did stop to support the old platform for projects, which was completed in May 2019. The project team is working on a new version of Disk Detective. The public is invited to search through images captured by space agency's Wide-field Infrared Survey Explorer and other sky surveys: the Two Micron All Sky Survey, the Digitized Sky Survey and the Sloan Digital Sky Survey. The images in Disk Detective have all been pre-selected to be extra bright at wavelengths where circumstellar dust emits thermal radiation. They are at mid-infrared, near-infrared and optical wavelengths. Disks are not the only heavenly objects that appear bright at infrared wavelengths; active galactic nuclei, galaxies, asteroids and interstellar dust clouds also emit at these wavelengths. Computer algorithms can't distinguish the difference, so it is necessary to examine all images by "eye" to make sure that the selected candidates are stars with disks, and not other celestial objects.
Classification
At the Disk Detective website, the images are presented in animated forms which are called flip books. The users view a flip book and classify the target object based on simple criteria, such as whether it is round in DSS2 or 2MASS images, or the object is extended beyond circle in WISE images, if there is more than one object in the circle or if it is off the crosshairs. The ideal target is classified as a "good candidate", and further vetted by the advanced research group into a list of "debris disk of interest" candidates. The selected disk candidates will eventually become the future targets for NASA's Hubble Space Telescope and its successor, the James Webb Space Telescope. They will also be the topic for future publications in the scientific literature.
The volunteers can join an exclusive group, called "advanced users" or "super users" after they have done 300 classifications. The advanced users then might compare the candidates with literature or analyse follow-up data. This advanced user group is similar to other groups that have formed in citizen science projects, such as the Peas Corps in galaxy zoo.
Discoveries
The project has so far discovered the first debris disk with a white dwarf companion and a new kind of M dwarf disk in a moving group. The project found 37 new disks and four Be stars in the first paper and 213 newly-identified disk candidates in the third paper. Together with AWI0005x3s, the Disk Detective project found 251 new disks or disk candidates. The third paper also found HD 150972 as a likely member of the Scorpius-Centaurus moving group, 12 candidates that are co-moving binaries and 31 that are closer than 125 parsec, making them possible targets for direct imaging of exoplanets. At the 235th meeting of the American Astronomical Societythe discovery of four new Peter Pan Disks were presented. Three objects are high-probability members of the Columba and Carina associations. The forth object has an intermediate likelihood of being part of a moving group. All four objects are young M-dwarfs.
The project did make estimates about the amount of high-quality disk candidates in AllWISE and lower-limit false-positive rates for several catalogs, based on classification false-positive rates, follow-up imaging and literature review. Out of the 149,273 subjects on the Disk Detective website 7.9±0.2% are likely candidates. 90.2% of the subjects are eliminated by website evaluation, 1.35% were eliminated by literature review and 0.52% were eliminated by high-resolution follow-up imaging. From this result AllWISE might contain ~21,600 high quality disk candidates and 4-8% of the disk candidates from high-quality surveys might show background objects in high-resolution images, which are bright enough to affect the infrared excess. The project also has a database that is available through the Mikulski Archive for Space Telescopes. It contains the "goodFraction", describing how often a source was voted as a good source on the website, as well as other information about the source, such as comments from the science team, machine learned classification, cross-matched catalog information and SED fits. A group at MIT did use the Disk Detective classifications to train a machine-learning system. They found that their machine-learning system agreed with user identifications of debris disks 97% of the time. The group has found 367 promising candidates for follow-up observations with this method.
