Google AI can identify wildlife from trap camera footage.

2019-12-18

Due to climate change, poaching, and destruction of natural habitats, the situation of some animal populations is much worse than others. It is estimated that the number of over 4,000 species has decreased by 60% from 1970 to 2014. A recent global assessment by the United Nations found that up to 1 million species are at risk of extinction in the next 10 years. To address this issue, Google has collaborated with international conservation organizations and other groups with the goal of helping to manage one of the world's largest and most diverse databases. This database is a collection of photos captured by motion-activated cameras and is part of wildlife observation. It also serves as a platform that supports artificial intelligence, simplifying conservation monitoring by accelerating the analysis of camera-trap photos. Tanya Birch, the project manager of Google Earth Outreach, explained in a blog post that biologists and land managers around the world place and monitor thousands of camera traps, which capture millions of photos each year. The sheer volume of these photos makes it challenging to effectively screen and identify animals, especially when they are in the dark or hiding behind bushes. Over 80% of the photos do not contain any animals at all, as they are triggered by wind or other irrelevant elements. The value of the Wildlife Insights dataset lies in its ability to filter camera trap images by species, country, and year. Google states that its AI Platform Prediction service enables the analysis of up to 3.6 million photos per hour by wildlife experts. This is 3,000 times faster than the average human expert, who can only label 300 to 1,000 images per hour. The models trained by Google can utilize its TensorFlow machine learning framework to classify species in the images and automatically remove photos without animals. It has achieved high accuracy rates for 614 species, ranging from 80% to 98.6% for species like the Amur tiger, African elephant, and rarer wildlife such as the white-lipped deer. With this data, managers of protected areas or anti-poaching projects can assess the health status of specific species, and local governments can utilize the data to formulate policies and conservation measures.

RECENT AI NEWS

RECENT AI TOOLS