The platform for interactive pre-processing camera trap

Service description:

The service will be able to (1) automatically filter out most (if not all) unwanted pictures from camera traps and (2) propose the species name. The idea is to create a Graphical User Interface that allows users to view the proposed species name. If they accept the suggested classification from the system, the picture is uploaded to any of the chosen biodiversity citizen observatories (iSpot, Artportalen, Natusfera, etc.) with the proposed identification name. 

Development & functioning :

  • Drastically cut the data analysed, archived or transmitted from remote camera traps. This will be achieved by:
  • Removing most images that are not useful (e.g. no animals present, near duplicates, false positives due to wind movement, etc.).
  • Providing compressions and filtering modes.
  • Provide access to robust state-of-the-art processing methods, algorithms, as a starting point for further analysis, comparisons, etc. This will be achieved by:
  • Providing automatic (possibly overlapping) regions of interest (ROIs), such as bounding boxes, for objects of interest (typically various animals).
  • Provide a minimum of automatic analysis of data and make the entire processing of images much easier by removing some of the algorithmic intricacies involved in cleaning up data, removing backgrounds, filtering, selecting useful representatives (high counts). This currently depends largely on the availability of high-quality, large-count training sets. This analysis can serve as a comparison gauge for other methods, algorithms (including human-based analysis). This will be achieved by:
  • Providing initial animal presence information: count, species presence/ID (when possible).

Innovation for citizen observatories:

The main problem for camera trap users is that they get hundreds (or thousands) of pictures that they (1) have to filter (some of them just captured the movement of the background plants/trees due to the wind, not because an animal appears) and (2) have to label properly. This service will address such problems.  There is no similar service (a) for camera-trap systems (partial solutions exist) and (b) on the cloud.
We are working on this service, it will be available soon!
35%

Keywords:

IMAGE DATA REDUCTION, COMPRESSION, FILTERING, ROI analysis, CONTENT ANALYSIS, COUNTING ANIMALS, BACKGROUND PROCESSING, GRAPHICAL USER INTERFACE

Coordinator:

Papers, documents and other experiences about co-designed digital services are collected in the following Zotero group