Protecting the Planet

Artificial Intelligence

Protecting the Planet

Green Goals: How AI could save the Planet

The steady growth in human population is putting increasing pressure on our planet, and we want to help protect it.

In line with the United Nations Sustainable Development Goals of sustainable cities and communities, life on land, and life below water, NVIDIA is developing technology that advances ways of sustainable and safe living in dense cities, protects food sources, encourages biodiversity, and monitors wildlife populations on land and marine ecosystems in the oceans.

GPUs Go Green

GPU computing — a driving force in advancing virtual reality, medical imaging, and autonomous cars — is also accelerating research focused on helping the planet.

Green-focused startups are literally in the weeds detecting plant diseases, in the jungles mapping rainforests, and on the savannahs helping to save endangered species. They are measuring drought, and in the oceans assessing the health of threatened coral reefs. They are improving the efficiency of solar-powered homes and recycling services.

Efforts to use GPUs in sustainability and other environmental projects align with the 17 sustainable development goals detailed by the U.N. to eliminate the world's most pressing problems — from hunger to gender inequality. These goals can be a boon to businesses, spurring new applications, sharpening thinking about supply chains, enhancing recycling endeavors, and reducing carbon emissions.

Digital Green Thumbs

Among researchers aiding farmers are Ecole Polytechnique Federal de Lausanne in Switzerland, which specializes in physical sciences and engineering, and Penn State University, who are using the Caffe deep learning framework and Tesla K40 GPUs to train a model that identifies crop diseases.

To share the information, the researchers created a website, Plant Village, an open access database of 50,000+ images of healthy and diseased crops. The goal is to launch a mobile app that lets farmers around the world snap a photo of their diseased plant and quickly diagnose it.

Silicon Valley-based Blue River Technology developed a deep learning solution called LettuceBot that rolls through a field photographing 5,000 young plants a minute, using algorithms and machine vision to identify each spout as lettuce or a weed. Using the Caffe deep learning framework, the company's neural network was trained with TITAN X GPUs to help make crop identifications and develop a “see and spray” technology so crops can be sprayed with fertilizer, and weeds with herbicide.

Research in the Amazon forest by Hannover, Germany-based PEAT Technology led to the development of the Plantix mobile app, which is helping farmers on three continents quickly identify plant diseases using AI. Farmers upload photos to the app, which runs them through image recognition software to spot pathogens.

The crowdsourced database grows more powerful with each new crop disease logged, helping farmers from Brazil to India identify, treat, and prevent crop diseases.

The health of tropical rainforests is drawing as much attention as farmed fields. AI and a powerful spectral imaging method developed by an ecologist at the Carnegie Institution for Science and Stanford University can map a rainforest in unprecedented detail.

By identifying each tree species by its chemical composition, the data shows the rainforest is more diverse than anyone thought. The map takes some of the guesswork out of protecting one of the most biodiverse places on Earth and could help in pinpointing new areas for conservation.

Deep Water Learning

Droughts parching regions as far-flung as California, Brazil, and Australia may be a preview of what's to come, scientists predict. With the prospect of available water diminishing, deep learning could help thirsty communities find out how much they have.

Orbital Insight developed a GPU-accelerated deep learning system that's designed to measure and monitor the surface levels of freshwater reserves across the globe based on images from the U.S. Geological Survey's Landsat 8 satellite, launched into space in 2013, which images the entire Earth every 16 days.

These images are fed into a neural network that searches for water, pixel by pixel, building a database of billions of pixels to train from. NVIDIA GPUs hosted in the Amazon Web Services cloud accelerated the training process.

Thanks to the Mountain View, California-based Orbital's mapping work, community water managers and policymakers can make better decisions about water management based on a clear picture of how much available.

Other watery regions, including coral reefs, also face threats – climate change, coastal development, overfishing, and pollution. With a quarter of Earth's coral reefs already gone, scientists are racing to save them using GPU-powered deep learning.

The world depends on the health of coral reefs, as they provide food and shelter for more than a quarter of all marine species, support fish stocks that feed more than a billion people, and provide jobs to millions of people in coastal areas.

A deep learning process using NVIDIA's CUDA programming model running on powerful GPUs automatically analyzes reef photos of corals, sponges, algae, and other elements. It's 900x faster than the traditional method but just as accurate, allowing scientists to quickly assess the health of reefs so they can take steps to protect them.

Recycling Robots

Recycling is becoming more sophisticated, safer, and less costly thanks to machine learning.

Barcelona-based Sadako has created a robot that uses machine learning to sort recycling and then extract specific types of valuable waste, like PET plastic bottles. A GPU-powered robot can differentiate types of trash and snatch selected items with a robotic arm.

The method can target specific kinds of trash for recycling, separating valuable materials from garbage, and relieving some of the burden on workers who traditionally sort through waste that streams down conveyors.