The Planet Four science team is collaborating with machine learning researchers in Australia. We’re working on a joint paper that looks at using the current Planet Four seasonal fan and blotch catalog that was generated from the HiRISE season 2 and season 3 images from the original Planet Four website. Michael Aye made some great images for the paper showing examples of what gets generated from your classifications. I thought I would share some of the figures.
In the figures below you’ll see the Planet Four subject image or tile on the left that we have asked volunteers to classify and the right the image is overlaid with the resulting fans (shown in green) and blotches (show in magenta) identified. Each Planet Four subject image has about 30 people review it and map fans and blotches they see in image. We then take all the individual marks and combine them together to identify which sources are fans or blotches. This blog post gives some more information on how we developed the clustering code that combines your classifications together.
With Planet Four 2.0, the current version of the project, we’ll be using the same method and software to combine the markings you are making now to identify the fan and blotches in the subject images. So think of the images below as a sneak peek of what your clicks will be turned into.