VultureWatch
The Problem
Vultures as a group are the most threatened birds in the world. Of the 23 species of vultures in the world, over 70% of them have an unfavourable conservation status, meaning that they’re at risk of going extinct. Poisoning is the biggest threat facing vulture populations in Africa, and in particular this threat has grown rapidly with the increasing slaughter of elephants by poachers. Poisoning associated with ivory poaching now accounts for one third of all vulture poisonings since 1970 and is currently the biggest cause of vulture mortality. This is also compounded by farmers purposefully doing poisoning to their own livestock in order to cause harm to predators. The current state of response is getting the rangers to find the vultures and prevent fatalities using posion response kits however there are no efficient means for the rangers to find these posioned carcasses before the Vultures are impacted
Our Proposal
The concept was to have a set of drones which would autonomously circulate between livestock farms (assuming knowledge of where these farms are, is known). Using a machine learning model trained on recognising carcases (goat and cow) the information from the drone cameras could be fed in and could flag these carcasses up to rangers. These carcasses would indicate that there has been poisoning and could allow the rangers to get to the area before Vultures descended.
We Assume that...
The main assumption is that there is enough Drone Availablity. The hope would be that as this technology gets cheaper and more ubiquitous (in Africa) that this solution could be rolled out if useful.
The time saved to reach the poisoned carcasses because of the drones would have a big impact on poison response.
There is a capacity to keep the drones flying without them being interfered with or need constant recharging.
These carcases are mostly not under a canopy/ shielded from aerial view.
The drones could survey an effective amount of small farms in the time of a full charge. The farms are extremely separated by distance.
There is a dataset that could be used to train the Deep Learning Model.
Constraints to Overcome
Getting information to rangers of where the poisoned carcasses are so that they can conduct a poison response where time is of the highest priority and reduce the spread of the poison to vultures.
Current Work
Testing assumptions of idea. Trying to obtain dataset for training.