Understanding Herd Mentality

As the world's human population grows and expands into previously wild areas, many animal species are facing a decline in population. This is due to a number of factors, including habitat destruction, pollution, and poaching. However, conservationists are working tirelessly to save these endangered species and prevent their extinction.

These efforts are widely seen as being critically important because ​​the loss of a species can disrupt the balance of an entire ecosystem. Each species plays a role in the ecosystem, whether as a predator, prey, or as a pollinator. The loss of one species can have a domino effect on the rest of the ecosystem, leading to a decline in biodiversity and potentially triggering the extinction of other species.

Declining animal populations can also have a negative impact on human society. Many species have important ecological and economic roles, such as providing food, medicine, or other resources. For example, the extinction of honeybees would have a significant impact on the world's food supply, as they are important pollinators for many crops.

For conservation efforts to be successful, detailed data on the behavior of large numbers of animals, and their interactions with one another, is needed. Without this understanding, attempts to help endangered species may be based on incomplete or incorrect information, and that can hinder these well-intentioned plans.

As is the case in many fields, high-quality data is difficult to come by. Traditional methods of tracking large groups of animals, like direct observation and biologging with instrumented tags, are limited in the number of animals that they can monitor. These methods are also restricted in the area of land and the amount of time that they can cover. For these reasons, it can be very challenging to understand complex patterns of animal behavior and social interactions.

There may be a technological solution to this problem, according to a team of researchers at the Max Planck Institute of Animal Behaviour. They have developed a drone-based system that can monitor large groups of animals using computer vision and machine learning. The system can keep watch over large geographical areas and for long periods of time. Moreover, it is able to very quickly provide highly detailed information about the location and behavior of each animal in the group in a way that would be virtually impossible for human observers.

The team flew a pair of DJI Phantom 4 Pro drones outfitted with cameras. The drones were manually piloted to keep all members of the herd under observation in view. The captured images were analyzed by a deep convolutional neural network that was trained to predict the locations of animals present in each frame of video. In order to track each individual animal across video frames, a modified version of the Hungarian algorithm was employed.

Another algorithm was developed to correct for the motion of the drone, and consider the effects of variable terrain on the location predictions. Next, a body-part keypoint detection method was used to collect finer-grained information about the activities of the animals.

It was noted that some animals did notice the presence of the drone, and that this factor did influence their behavior in certain cases. The drone was given away primarily by the sound of its motors, so the researchers suggest selecting a quieter drone model, approaching from downwind, and flying at higher altitudes may be effective ways of preventing disturbances to the target animals.

The methods outlined in this work are applicable to many species of animals, so the team hopes that it will prove to be a useful tool for researchers across many fields. As others implement similar systems, the high resolution data may reshape the way we approach conservation efforts for the better.