Images as sources of intelligence

CAIMAN IS UNIQUE

CAIMAN is the first truly integrated camera-trap solution. All classified pictures are stored in your own private database, that is also used to store all your other data, including Cluey-observations, tracks, spreadsheets, map layers, and other data that you’ve uploaded to WITS. From there, you can use Focus and the WildCAT-apps to combine and analyse data and create maps and reports, including species range maps, statistical overviews, and more.

SAVE VALUABLE TIME

In many wild areas, camera traps are placed to obtain critical data about wildlife and their habitats. Camera traps are significantly more effective than other methods at detecting a large number of species and for generating detections of (rare) species. The traps are placed on wildlife paths and a photo is taken when movement is detected.

Save manpower: in the past, all images were manually checked by rangers or biologist and labelled with the species captured on the picture. However, these are hours and hours of manual labor that these experts could put to much better use. Especially since more than 50% of all pictures taken do not contain animals, but have captured “waving grass”.

Reduce latency: information is most valuable when it’s hot. If a rare species has been detected, you want to know it A.S.A.P. By automating the classification of images, results are available within a coffee-break.

This was originally posted on the Sensing Clues website.

CAIMAN

CAIMAN is our solution to automatically classify animals on images from camera traps.

C= camera trap
AI - artificial intelligence
MAN = the (hu)man in the loop

We cannot only detect which species are on the picture but also how many of each species. The models that we develop belong to the best of world.

How can we be so sure? Well, we came second in the worldwide Kaggle iWildcam competition of 2021.

HOW DOES IT WORK?

Everyone can upload their photos from cameratraps to our server. All photos are sent through our artificial intelligence (AI) models and the following information is added to each picture:

• is there an animal on the picture?
• which animal is on the picture?
• what is the certainty rate of the classification?

If the confidence percentage is too low, we send the photo to a website where rangers and biologists can check the classification and change it if necessary. This is the “Human in the Loop”.

FIRST RELEASE

The first fully operational version of CAIMAN is expect to be available by the end of 2021. If you want more information or want to participate, don’t hesitate to contact us.

This was originally posted on the SensingClues website.