.png)
About Gardyn
Grow Fresh and Healthy, All Year Round.
Industry
AgriTech, Hydroponic Farming, Aeroponic Farming, Artificial Intelligence, Software
Headquarters
8120 Woodmont Ave., Bethesda, Maryland, United States
Gardyn is a fast-growing pioneer in homegrown food and a leader in vertical hydroponic gardens that aims to reinvent providing everyone access to fresh produce year-round. Free of pesticides and food waste, Gardyn is a more sustainable and healthier alternative to produce that is often harvested from far away and weeks from making it to store shelves. Gardyn has been awarded Best Invention of The Year by TIME Magazine and Sustainability Product of The Year by Business Intelligence Group. Using innovative hybriponicTM technology, a patented blend of hydroponic and aeroponic farming, Gardyn has proven itself as a revolutionary company by setting a new standard with at-home technology that requires 95% less water than traditional methods.
Gardyn’s Challenge
Gardyn uses proprietary technology to analyze growth stages in plants and aims to optimize plants’ growth using AI-based techniques. Coupled with their innovative hybriponicTM technology, they monitor individual plants’ water cycles and visible growth using two on-device cameras that take photos every half hour. Watering patterns are optimized using AI to boost plant's germination and growth.
Scaling with technology is fundamental for Gardyn as manual analysis can be difficult and cumbersome for Gardyn’s growth optimization team. In order to streamline the process of growing plants at home, Gardyn seeks to minimize the necessary human intervention so that plant maintenance does not become a burden for the user. As such, Gardyn wanted to develop a scalable, automated computer vision solution to smooth the growing journey for users.
Datature’s Solution
Gardyn has developed an AI-powered platform, Kelby, that is designed to automatically monitor plant growth and device health based on image capture from the cameras along with data coming from other embedded sensors (temperature, humidity, water level, etc.).This translates into minimal, efficient, and actionable recommendations and alerts for users to understand how to interact with their Gardyn.
To develop Kelby, Gardyn leverages Datature’s end-to-end no-code platform to develop production level instance segmentation models, as well as construct a streamlined, automated MLOps pipeline to create a sustainable computer vision model life cycle.
At the image annotation stage, Datature's Nexus allows Gardyn to manage the annotation workflows for a team of annotators through Datature’s Automation tab, as well as provide an on-platform Annotator to speed up the annotation process. Datature's platform enables the entire team to collaboratively annotate and review annotations through distinct user accounts, significantly expediting the annotation and quality assurance procedures.
With the large influx of raw image data everyday, Gardyn requires tools that facilitate efficient and organized annotation. With Datature’s Automation tool, the team is able to use asset group tags to control which images are to be annotated, assign specific collaborators for annotation and others separately for review, as well as pathways for fixing or marking annotations as complete. All these complex logical workflows are made easy to control through a modularized interactive interface, where Dataset, Annotate, Review, and Completed nodes can be created and connected to construct a clear visualization of the intended annotation workflow. Gardyn can efficiently inspect and modify their previously generated labels while also generating new ones rapidly and accurately with intelligent labelling tools such as IntelliBrush through the Annotator. Gardyn annotates images to recognize crucial features of the different growth stages in order to precisely and efficiently determine the current growth stage of a plant, as well as determine whether the pods are empty or available for more plants to be grown. At the Review stage, reviewers can check annotations and reject or approve them for further fixing or completion. To keep track of the progress at each stage of the workflow, the Automation page also provides basic statistics about the aggregated statuses of the images at every node in the workflow.
After the labelling process, the team at Gardyn was able to easily and rapidly train a robust instance segmentation model that has proven its effectiveness in terms of speed and accuracy at the production level for the past year. This agile development was enhanced by the ease of experimenting with different training setups concurrently as well as the various visual aids to help users understand the impacts of their various trainings. Gardyn leverages the traditional loss monitoring training dashboard as well as the evaluation preview function under Advanced Evaluation, that displays predictions on sample images at each evaluation checkpoint as the training is ongoing. This helps Gardyn’s developers observe how well the model performs and if it is improving both quantitatively and visually at various checkpoints throughout the model training process.
At the deployment level, Gardyn’s Kelby leverages Datature’s Inference API to reduce the development necessary for a production-level deployment to a no-code deployment setup on Nexus and just a simple API request to get rapid predictions on images captured in real-time rapidly. Datature’s Inference API provides fully maintained, dedicated GPUs on the cloud to host Gardyn’s robust segmentation models to enable Gardyn’s team to freely retrieve predictions with a widely supported tool like an API request, and allows the team to focus on actionable insights based on the predictions rather than the tedious and complex setup needed for a local deployment or setting up their own cloud deployment.
To facilitate iterative development and a more sustainable and efficient lifecycle for the model, Gardyn’s team also utilizes Datature’s Management API and Python SDK to automate processes and reduce latency between major steps in the MLOps, such as automatically uploading images for labelling, starting a new model training when images have been fully labelled, and redeploying the latest models to ensure that the model deployed on the Inference API is providing the most accurate predictions.
Overall, Datature’s tools have allowed Gardyn to not only develop an effective computer vision solution to automatically detect plant growth stages and other useful visual information for actionable insights, but also sustain a sophisticated and efficient pipeline to continuously improve on the model and overall user experience.
“Datature’s platform has allowed us to scale our AI efforts across the board. Their slick UI and deep feedback on model performance allowed us to deploy production-grade segmentation and detection models without a large team or major expense. The Datature engineers provided us with invaluable guidance and support at every stage, and we’re now expanding our use of the platform to involve internal and external labelers. As the use of AI inside Gardyn is scaling up, Datature is providing the tools to do that without a hitch.” - Sunil Rawal | Lead AI Engineer at Gardyn
Datature is proud to support Gardyn’s mission to optimize and automate plants’ growth and we are excited to see what’s next for Gardyn.
If you are an innovator looking to unlock the potential of deep learning to transform the agritech industry, get in touch with us!
About Gardyn
Gardyn reimagines the future of food. Gardyn develops cutting-edge technologies to make it possible and convenient for anyone to grow large quantities of nutritious and tasty produce at home, in a fully automated way, within just two square feet and no green thumb required – fully local, highly sustainable. After several years of research and development in partnership with universities, Gardyn launched the revolutionary “Gardyn Home” device and its innovative gardening assistant KelbyTM, which won several awards and broad media attention.
About Datature
Datature is an end-to-end MLOps platform that allows teams and enterprises to build computer vision models without a single line of code. Teams can manage datasets, annotate, generate synthetic data, train and deploy - all in a single, secure cloud-based platform. With the rise of citizen data scientists, deep tech companies, and enterprises looking to adopt deep-learning - Datature equips these startups / experts with the tools required to build their own capabilities easily within weeks.
Build models with the best tools.
develop ml models in minutes with datature