News Categories

Conservation AI, with the help of NVIDIA, detects real-time threats to endangered animals

By The Count - on 21 Feb 2023, 12:00pm

Conservation AI, with the help of NVIDIA-based solutions, detects real-time threats to endangered animal species

Conservation AI, a United Kingdom-based non-profit conservation organization, is making use of NVIDIA technology to analyze satellite imagery, and other sources of data to uncover potential threats to endangered animal species in real time.

The organization makes use of its Conservation AI platform, which employs more than 70 edge AI cameras, to capture imagery and other metadata of animal species that are under threat of extinction.

(Source: Screengrab from Conservation AI video)

The Conservation AI platform also comprises NVIDIA Jetson modules for its edge computing requirements, and the NVIDIA Triton Inference Server offers backend support for the machine learning framework. The system analyzes footage “in just four seconds”, detects species of interest, and sends out email alerts to researchers of “potential threats”. The platform’s edge computing environment leverages on the NVIDIA DeepStream framework for its video analytics requirements.

Conservation AI aims to protect endangered animal species through real-time detection of threats that are based on historical data. The imagery and other metadata collected from its edge AI computing environment provides trained models that can be reused and applied on its Triton Inference server-hosted frameworks. This is achieved through transfer learning on the said NVIDIA open-source Triton model repository.

According to Conservation AI, a typical camera trap study is too time-consuming for data analysis; however, with the new platform, the organization is able to analyze drone footage in real-time so that interventions can be implemented without delay.

The platform also allows conservationists to speed up deep learning inference by up to four times, compared to the previous solutions employed by Conservation AI. The deep learning models are trained with a collection of NVIDIA Turing-based GPUs that comprise the Quadro RTX 8000-series, and the Tesla T4, as well as the Ampere A100 GPU.

One of the challenges faced by conservationists is the lack of actual imagery to train AI models; as a result, the conservation organization is also exploring the use of synthetic data for model training. Ultimately, the conservationists hope to make use of NVIDIA’s AI-powered computing stack to reverse the loss of biodiversity in the face of challenges posed by climate change and destructive human activities. Read more about the forward-looking initiatives of Conservation AI here.

Source: NVIDIA

Join HWZ's Telegram channel here and catch all the latest tech news!
Our articles may contain affiliate links. If you buy through these links, we may earn a small commission.