Introducing wildthing

CEI has developed a revolutionary AI and Machine Learning program for Counting and Tracking Wildlife from Drones and UAVs. Some of our results are shown below. R & D is continuing to refine our accuracy and increase our number of species in the identification model. The model should be finished and deploy able within two months.

Species Identification, Counting and Tracking system for Wildlife using a UAV


The study objective is to develop a field tested and deploy able UAV system to identify, count and track wildlife species endemic to Alberta Canada.

pROJECT Description

UAV technology has developed to a state that off the shelf systems can be used to acquire multiple sensor data over large enough areas and a low enough cost to allow them to be used for detailed wildlife monitoring. Object recognition software using artificial intelligence, machine learning and deep neural networks has progressed at an even faster pace so that now well equipped programmers and scientists can train this software to identify virtually any object or in this case wildlife species. To date little effort has been put into integrating these two technologies into a wildlife identification and tracking system. CEI through the use programmers and trained UAV pilots will develop a software package and computer platform that can be deployed in a small (<10 kg.) UAV. This package will have a Graphic User Interface (GUI) that is intuitive and user friendly and can be trained to recognize a variety of wildlife species endemic to Alberta, Canada. The system will then be field tested for accuracy of species Identification and counting at the CEI facilities. This revolutionary package will then be made available to the Environmental Community for training and deployment anywhere these services are needed.


The project parameters are as follows:

• The primary objective for the project is to develop a software package in a Windows 10 environment that can analyze sensor data (HD video, HD stills, Thermal and multi-spectral data) collected from an UAV..
• The analysis can be real time on board the UAV with results fed to a ground station or data only collected on the UAV and analyzes done at the ground station in real time or after the flight.
• Flight time for the UAV should be in the area of 2 hours.
• Software should be based on either Detectron2, YOLOVO4 or similar object recognition software and must operate through a GUI.
• Software must be able to be trained to recognize species of mammals and birds endemic to Alberta.
• Software training will be done at Cochrane Ecological Institute (CEI)
• UAV and software will be field tested at CEI property.


Example of HD Video and Buffalo

Example of Thermal Imaging video (FLIR) of Buffalo same image as above