Data from "Modelling calibration uncertainty in networks of environmental sensors"

A set of (slow motion) videos of bumblebees taken around 2020. Dataset also contains a CSV file with citizen science guesses as to the species in each, guesses by a neural network and a ground truth (from an expert).

Videos

The videos (are stored in the set of zip files, to make it 

easier to download/upload.

Videos: 

 - taken using smart phones with slow motion mode.

 - codec: H.264 (High Profile).

 - 1280 × 720.

 - mp4.

CSV

The CSV file contains the following headers:

  > video -- the filename of the video (see in zip files)

  > person 1..person 5 -- the guesses by each individual as to the species. No entry means they didn't guess for that entry.

  > CNN -- the guesses by the convolutional neural network*

  > ground truth -- the guess by an expert.

Most entries are either blank or the scientific name for a species. There is also 'whitebuff' -- which means that bee was either Bombus terrestris or one of the Bombus lucorum species, but are very often too difficult to distinguish in the field, so have been combined.

* For how these predictions were made, see Ollett, J. (2021) Bumblebee classification with convolutional neural networks. Undergraduate final year dissertation, University of Sheffield.