.svg){kind=small}
.svg)
.svg)
Lynx: AI-Powered Computer Vision for Bat Conservation Automating wildlife population monitoring to protect critical ecosystems through high-precision, low-light video analysis.
Lynx – AI for Bat Conservation: Saving Ecosystems, One Wingbeat at a Time
Bats are vital to our ecosystems: they pollinate plants, control pests, disperse seeds and reduce disease spread. One bat can eat hundreds of mosquitos in a single night, reducing the spread of disease like Dengue Fever, Malaria or Chikungunia.
Yet, over 22% of bat species in Europe are threatened due to habitat loss, climate change, and human activity. Accurate population monitoring is critical for their conservation—but current methods are painfully slow and imprecise.
Naturalists spend hours manually counting bats frame-by-frame in low-light videos, a process prone to errors and unscalable for large colonies that sometimes count thousands of individuals.
This is where Lynx steps in.
Lynx is an AI-powered SaaS platform that automates bat counting with >95% accuracy, using a cutting-edge computer vision methods. Unlike manual methods or outdated software, Lynx processes videos 100x faster, turning hours of tedious work into minutes of actionable insights. Researchers and conservationists can now:
Monitor large colonies in real-time, even in low-light conditions.
Why Lynx?
Speed: few minutes per video vs. hours or days manually.
Accuracy: >95% detection rate.
Accessibility: No installation—just upload videos and get results.
Scalability: From small research teams to national parks.
The Impact
By automating bat counting, Lynx empowers conservationists to act faster. In Reunion, France, our pilot project estimated the population size for the first time one of the largest colony of the island with data that would be impossible to gather manually.
With Lynx, we’re not just counting bats; we’re saving ecosystems.
Provide a fast and standard way to monitor bat population.
We use a combination of optical flow, object detection and object tracking models to be able to count bats. We fine tuned pre-trained models with our own dataset, made of collected records on the fields and given by our partners. To make the annotation faster, we developed a first model using traditional computer vision methods to get basic annotations, then we performed manual annotations to clean the training dataset. To make inference execution fast we quantized our model.
Divide the time needed to count bats per 100.Only one ecologist needed for a whole site, with one camera per colony exit point. While without Lynx we need one ecologist per exit point.
We show that technology and sustainable development can work together. Sometimes people think technology is always bad for the environment.
.svg){kind=icon}