The GDELT Project

Deep Learning Image Identification To Counter Poaching

One of the most powerful aspects of "deep learning" (neural network) image recognition is its ability to identify particular species of animals and broad concepts from novel imagery. Using the Google Cloud Vision API we ran some experiments this afternoon over the last 48 hours of news monitored by GDELT to identify a sample of articles from the GDELT Global Knowledge Graph that had been assigned a Theme of "Poaching" and also contained an image identified by the Google Cloud Vision API as being of an elephant, rhino, or fish/fishing.

The collages below show a sample of the images returned for these searches, illustrating the tremendous potential of this approach.  Notice how in the case of the fishing search, the Cloud Vision API correctly identified photographs of fishing boats, allowing broad concept-based searches.

ELEPHANTS

RHINOS

FISHING