With earth’s population swelling and worries of food shortages a growing concern, scientists and researchers around the globe are tapping GPUs to solve pieces of the food puzzle.
From livestock and crop care to dietary assessments, GPUs are helping to speed up the training of deep-learning computing models needed to tackle the problem at unprecedented scale.
For instance, Connecterra, an Amsterdam-based startup, has been working on a GPU-powered deep-learning platform for monitoring the movements of livestock.
The company’s technology consists of a wearable device that transmits real-time information on a herd to a cloud-based platform for analysis and prediction of behaviors.
Connecterra has been using a GeForce GTX 970 GPU to train its deep learning network. This will provide increasingly detailed insight about livestock tendencies, enabling ranchers to improve production and optimize breeding patterns.
Where the Deer and the GPU-Powered Robots Roam
On the other side of the planet, a team at the University of Sydney is working on a solar-powered robot, dubbed SwagBot, that will play the role of high-tech cowboy to help monitor livestock.
SwagBot will roam among grazing cows, horses and sheep, herding them at times, while also providing video and other data that will help ranchers improve operations as well as animal health.
Currently focused on the drive train and articulation of the robot, the team will soon move on to developing autonomy for the robot’s software and numerous sensors, said project leader Salah Sukkarieh, professor of robotics and intelligent systems.
Sukkarieh said his team is using the GeForce GTX TITAN X for offline training of machine learning models that will support SwagBot’s autonomy.
GPU-trained machine learning models also are being used to protect crops. Silicon Valley-based Blue River Technology uses the GeForce GTX TITAN X and the Caffe deep learning framework to train its LettuceBot to distinguish between lettuce and weeds.
The idea behind LettuceBot is to help crop farmers combat converging trends: the increasing resistance of weeds to herbicides, and the decline of available chemical treatments.
LettuceBot is already being used in fields that provide 10 percent of the lettuce supply in the United States. Blue River estimates its technology can help farmers reduce chemical use by 90 percent.
Data: Fresh from the Farm
While Blue River helps to protect healthy lettuce, a trans-Atlantic effort aims to detect and prevent crop diseases with help from GPUs.
Researchers at Penn State University and Switzerland’s Ecole Polytechnique Federale de Lausanne are tapping the power of NVIDIA Tesla K40 GPU accelerators running on EPFL’s Deneb cluster to train a deep convolutional network that’s learning to identify crop types and diseases from smartphone images.
Farmers around the world submit the images via the research team’s PlantVillage mobile app, which aims to make food-growing insight universally available.
Big Data, Healthier People
Of course, all of these efforts to ensure the health and nutrition of our foods, not to mention the efficiency of the farmers who raise and grow it, will mean less if we’re not eating healthy.
Along those lines, a team of researchers at the University of Massachusetts Lowell is attempting to aid in dietary assessment — and help combat the obesity epidemic — by ensuring accurate assessments of the calories people consume.
Their work uses deep learning to recognize the types and portion size of food to improve the accuracy of calorie-intake reporting. Four Tesla K40 GPU accelerators are helping train their food recognition algorithms.
With GPUs powering everything from cattle-monitoring systems to calorie counting, it only makes sense that they should also help in the preparation of food.
Luckily, we’re also covered there, with the amazing June Intelligent Oven. The oven uses deep learning techniques and an NVIDIA Tegra K1 processor to prepare your food to perfection, every time.
So, the next time you’re eating a delicious, healthy meal, give a toast to the GPU.