OS i biodling!

 In pr, utställning, vetenskap

Vid Apimondia Montreal 8-12 september träffas alla världens biodlare, biodlingsindustrin och forskningen för att utbyta erfarenheter.

BeeScanning deltar med utställning, monter C165, och föreläsning. Välkomna, hoppas vi ses där!

Abstract här:

BeeScanning – monitors bees health

Knowledge when needed to discover parasite varroa, dwv, queen and colony strength. All in your camera.

With a smartphone using the camera and the BeeScanning app, beekeepers can instantly diagnose the parasite mite varroa destructor as well as other features in the hive. Images are taken on living bees on the comb. The tool will also form basis for selection in breeding programmes as well as the basis for population modelling research in the colony. New features are added including brood diseases and bee behaviours, evolving technology using video and sound. Enabling a tool for enhanced knowledge.

BeeScanning is funded by the European Innovation Program, the Swedish Board of Agriculture, The Swedish Agency of Innovations, via Kickstarter and 10 national and international awards.

The app is free to download and free to trial. For now there are 3 600 users, 20 000 images and more than 14 000 manually annotated regions in 15 classes. View http://tagger.beescanning.com Analysed images are stored and used for furhter training the AI. Images are instantly analysed in the following categories :

  • Bees with varroa 
  • Queen
  • Deformed wings
  • Number of bees 

Mean average precision: 83% Recall 67%

  • Results are presented to the user in absolute figures and as mean varroa infestation level %.
    Found objects are labeled and displayed to the user for review. Histograms indicates trends of the colony performance as a tool for decisions on treatment and/or breeding purposes.

The Beescanning technology is based on proprietary convolutional neural network, NN, and deep learning. Not by classification or algoritmic image analysing. This means results will continue to improve as the artificial intelligence learns from the ever increasing data its fed by the users.

Metadata as time, geopostion, breed is collected from growing the worlds largest database of images of bees on combs. We aim for a tool that can monitor events, nutritional status, health and make prognoses.

BeeScanning is based on our findings that there is a correlation between the actual varroa infestation level, as measured by alcohol washing or Apistan, and varroa that is optically detectable. https://beescanning.com

Björn Lagerman, founder Fribi Holding AB

Recent Posts
%d bloggare gillar detta: