L.A. Twitter

CODE: GitHub
DATATYPE: images
SOURCE: Twitter
PLACE: Los Angeles
YEARS: 2011-2014
SIZE: 100,000 images; 1M slices
SUBSET: random sample
VIZTYPE: slice histogram
FEATURES: hue, brightness
FEAT SRC: scikit-image
BINS: 1440
MORE INFO: Software Studies Blog
GALLERIES: ZKM Karlsruhe SAP Germany

This is a "slice histogram", an image histogram using slices of images. This is a hue histogram sorted vertically by brightness. The rationale for using slices instead of whole images is that image slices have lower standard deviation of their color properties, and so can produce visualizations that better reveal the color properties of the data (same would apply to other local visual features as well). One problematic feature of these visualizations is that often certain bins will be very tall; here, we've had to crop the plot so it fits to a square. This problem is solved with the flat histogram, which we started using in Winter 2016. It can also be solved by making the columns double- or triple-wide.