Creation of an R package for processing of image annotations

Context summary

In the context of participatory sciences, Ifremer has developed a video game for the general public to annotate images of the deep sea: Deep Sea Spy. Non-expert players are led to recognize different animal or plant species on marine images. The annotations used by the players to identify individuals differ according to the species: they can be points, lines, or polygons. The objective is to analyze the annotations collected. In particular, the aim is to detect which individuals are observed by multiple users, how many animals/plants of each species there are, and the size of the species.

Our intervention

• Deciphering project issues, facilitated by ThinkR’s in-house expertise in fisheries and mapping tools.
• Use of spatial analysis and mapping tools to process data
• Proposal of indicators of position quality, detection accuracy
• Taking into account the different types of objects detected in the analysis (points, lines, polygons)

Result & added value

• An R package with tools that allow researchers to detect the different annotations, group them together and define indicators.
• A documentation of the package for researchers to restart analyses if necessary.
• An innovative project that diverts the usual use of mapping tools (cf SatRdays 2019 presentation)