Taranis is a precision agriculture intelligence platform that helps farmers to monitor their fields and make decisions to reduce crop yield loss. Taranis uses Google Cloud Platform to enable high-volume drone image uploads and train TensorFlow machine learning models, helping farmers reduce crop loss and feed the world’s increasing population.
Eli Bukchin, Co-founder and CTO
"Agriculture is a seasonal business, so we have certain months of peak activity followed by quiet months, and we also have peaks throughout the day. During quiet times, we can scale back all our high-level Compute Engine GPU resources automatically so we don't have to prepare our system in advance."
“We use drones to take super high-resolution photographs of fields to leaf level, in addition to using satellite and plane imagery,” explains Eli Bukchin, Co-founder and CTO at Taranis. “The big problem farmers face is maintaining oversight on hundreds of thousands of acres: up to 40 percent of crops are routinely lost because of insects, crop disease, weeds, and nutrient deficiencies. We have a team of 30 agronomists who work on tagging images to feed AI models that enable us to identify problems before they affect the crops, so farmers can intervene earlier in a more targeted way and use fewer chemicals.”
As Taranis works all over the world and often in remote locations, it needed to find a way to upload large volumes of image files, along with creating a scalable infrastructure powerful enough to test complex machine learning models. Google Cloud Platform (GCP) along with TensorFlow provided the answer.
“We collect vast amounts of data from all over the world, in places with not much connectivity including Russia, Eastern Europe, and South America,” says Eli. “Developing methods to upload data in those conditions is a challenge. Our priorities are better connectivity and speed.”
What we did
“For our machine learning model training pipeline, we use TensorFlow,” Eli explains. “Choosing TensorFlow has helped us develop our models rapidly, as there is a lot of support available through the open source community. To develop our models, we use tens of millions of photographs that we have collected over the past year and a half, which have been analyzed and tagged. Each photo might have up to a thousand items of interest, such as insect damage or leaf discoloration, so the data volumes are really significant. In total, we have processed around 100 million distinct features in around 700,000 images.”
– Builds machine learning models using millions of tagged image data points, enabling disease, pest, and abiotic stress detection.
– Supports faster uploading of drone images in remote locations, reducing upload times from a day to several hours.
– Supports a scalable data pipeline with Compute Engine, Kubernetes Engine, Cloud SQL, and Cloud Pub/Sub, helping to improve stability
Get in touch
Looking for a cloud partner? We’ve got you covered.
Our team of experts can take care of your cloud, for much less than the cost of hiring that talent in-house.