Welcome to AI for Coyotes
A smart solution to co-exist with the "Song Dog"
Coyote (Canis latrans), the talking dog, is a species of canine, native to North America. They are revered as “God’s dog”, by Navajos, the Native Americans of the Southwestern United States. They are wild dogs that weigh from 25-35 lbs in the western region but can reach upto 50lbs are larger in the North Eastern region. They have pointed ears and a narrow muzzle. Their coat color ranges from gray to shades of red, depending on geographical location, but irrespective of their color differences, all coyotes have a black-tipped tail.
Coyotes are native only to the western two-thirds of the USA, but are currently found throughout North America from Alaska to as far south as Costa Rica. They are extremely adaptable and live in a wide range of habitats including urban environments. Coyotes are crepuscular - more active in the twilight hours. They have acute hearing and sense of smell, but they are color blind with less developed sight. Coyotes are called the "Song Dog" for a reason. They are the most vocal of all North American wild mammals. They use a panoply of vocalization sounds to communicate with their species, out of which 3 are the most distinct ones - the squeak, distress and howl call. Coyotes are monogamous, which means they often have one mate their whole life. They are very intelligent and devoted caregivers.
Importance of Coyote for the Ecosystem
Coyotes are a keystone species, meaning that their presence or absence has a significant impact on the surrounding biological community. By exerting a top-down regulation of other species, coyotes maintain the balance in the food web below and around them. When coyotes are absent or even just greatly reduced in a natural area, the relationships between species below them in the web are altered, putting many small species at risk.
Coyotes have existed in North America since the Pleistocene, and is the most persecuted native carnivore in North America. An estimated half a million coyotes are killed every year in the U.S.- one per minute. Coyotes are killed using methods such as the setting of traps and snares, the use of poisons, aerial gunning, and denning, for many or no reason, but, mostly to safeguard livestock. But, environmental studies and research show that all the killing just gives you more coyotes. Coyotes are pack animals consisting of alpha pair (dominant male and female) with betas and omegas (extended family members). Only the alpha pair breeds, producing one litter a year. When the alpha pair is killed, the subordinate pack members breed and bear larger litters of bigger pups. Researchers also found that in heavily hunted populations, adult survival rates drop, but, more yearlings start to reproduce with an increase in litter size. Thus, the increased persecution of coyotes leads to bigger populations!
Our premise is that most coyote-human conflicts can be avoided if people can be forewarned about the presence of the coyotes in time so that they can take preventive measures. For example, if people know that a coyote is headed towards their homes, they can bring in their small pets and secure their livestock. The alerts must be accurate and timely so that such protective measures can be taken.
For humans, avoiding the loss of valuable domesticated animals takes away the motivation to attempt to eliminate coyote populations. Likewise, if coyotes do not find easy prey, they are not likely to have the strong motivation to come into human habitations. Also, if humans know that coyotes are in the area and can peacefully co-exist, then the strong motivation to remove them will not be there.
We see that in most areas with human activity, mobile phone service is available and people carry their mobiles with them all the time. Therefore, mobile phones would be the best way to send alerts in a timely fashion. We also find that advancements in Deep Neural Networks have made it possible to identify specific animals in images. We have developed a system with a two stage Deep Neural Network, we call CoyoteNet, that uses images from home surveillance cameras or low cost cameras placed near typical coyote paths to detect the presence of coyotes. The first stage in CoyoteNet is an object detection system using the SSD MobileNet model in TensorFlow that detects any animal in an image. This is followed by our custom Convolutional Neural Network classification system that identifies coyotes from the animals detected in the first stage. CoyoteNet is optimized for low power and low memory so that it can be embedded in edge devices such as surveillance cameras or low cost special-purpose cameras. The components surrounding CoyoteNet can alert those living near the specific location where such detection occurs via their mobile phones, using Cloud based notification services. We are committed to licensing our technology free of cost to any company that wants to embed it in their products to protect coyotes.