On December 17, 2019, Google’s Earth Outreach team and six conservation organizations launched Wildlife Insights, a cloud platform that uses AI to sort the flood of images produced by motion-triggered camera traps. The partner organizations were the Wildlife Conservation Society, the Smithsonian Conservation Biology Institute, the North Carolina Museum of Natural Sciences, WWF, the Zoological Society of London, and Conservation International, and the launch made public the largest collection of camera-trap records assembled to that point.
Camera traps are central to modern wildlife monitoring, but they generate a crushing volume of useless frames: up to 80 percent of photos contain no animal at all, triggered instead by wind-blown grass or shifting light. Researchers traditionally reviewed every image by hand. Wildlife Insights trained a custom model on roughly nine million images contributed by the partners, running on Google Cloud’s AI Platform, and could classify images up to 3,000 times faster than a human - about 3.6 million photos an hour - discarding empty frames and labeling species automatically.
By collapsing months of manual review into near-real-time analysis, the platform let conservationists track population trends, detect rare or elusive species, and assess habitat recovery, as demonstrated after Australian bushfires. The project sits alongside Google and conservation efforts to standardize and share biodiversity data, and the underlying model later evolved into open-source releases for the wider community.
Why business readers should care: the value here was not detecting animals but discarding noise - automating away the 80 percent of empty images freed scarce expert time for the work only humans could do, a pattern that recurs whenever AI is bolted onto a high-volume sensing operation.