We worked with RESOLVE to put their data science capabilities into high gear – deploying over 20 smart camera applications in 6 months across Asia and Africa. By creating AI using only synthetic data, we were able to improve their operational efficiency by 2000%, increase product battery life by 42x, and develop object detection algorithms for animals that are rarely photographed in the wild.
The pain points
RESOLVE is an international environmental NGO based in Washington DC. They are the developers of TrailguardAI – intelligent camera systems that tackle worldwide environmental issues. Their requirement when they came to CVEDIA was to create solutions that track endangered species in the wild, and alert authorities of poachers and illegal traders. TrailguardAI is a camouflaged edge device that sits overhead in trees and captures images of species and intruders.
As an annual $175 billion illegal industry, countries across Asia and Africa are fighting to find solutions to wildlife trafficking. RESOLVE knew that this type of technology could have a massive impact on the illegal trade industry, but being endangered, few photos and videos of these species exist – making it nearly impossible to create effective detection models through traditional data collection. To extend the problem, the current cameras had to be changed out every two months due to AI power restraints.
RESOLVE was stuck at a stand-still – they knew exactly how they wanted their cameras to perform, but without effective algorithms and a better battery life, their cameras would send false positives to authorities too often and be a burden to maintain.
CVEDIA creates custom computer vision models from synthetic data. When RESOLVE came to us, we scoped out a project timeline for one one of their first use cases – an elephant detector. The CVEDIA data science team has developed a close to automated AI training system, so most of the time allocated to the project was allocated to creating custom-built 3D models of elephants. Our art team worked with their biology consultants to create life-like, photorealistic 3D models to train with.
By combining simulation with our data science process, our development team was able to recreate possible scenarios the cameras might pick up in their natural habitat. Within three weeks, RESOLVE’s custom solution was built – an intelligent camera application that flags elephants as they cross into the field of view, and takes a photo. Our hardware engineering team also worked to optimize the system’s performance, bringing the battery life up to an expected 7 years.
"Now typically we want cameras to last a year, even two years. You don’t want to have to go change the battery every two months, but in fact, that’s what we had to do. So the experiment we just did in South Africa, a typical rate of intruders (at a given trail) is maybe between one to ten transgressions a month. If we do the math... that should be the equivalent of 7 years of use in the field."Director of Biodiversity and Wildlife Solutions, RESOLVE
In under three weeks, RESOLVE went from concept to scale with their first intelligent camera deployed with CVEDIA technology. Within a month, RESOLVE had an intelligent camera running a CVEDIA algorithm deployed in Asia, capturing photos of wild elephants for research with geologists, conservationists, and local communities.
RESOLVE plans to deploy their cameras in 200 parks within a year. Through this creative use of technology, RESOLVE has achieved a breakthrough against illegal wildlife poaching, allowing them to grow their donor list and fundraise with an active, in-the-field solution.
CVEDIA has since then built over 20 AI systems for RESOLVE, for endangered species including snow leopards, rhinos, and tigers. RESOLVE aspires to monitor every square mile of wildlife on the planet with their solutions, effectively eliminating environmental crime. Watch the video above to hear Eric Dinerstein, Director of Biodiversity and Wildlife Solutions at RESOLVE discuss his experience with CVEDIA.