Tomato Imaging System
UX Designer @ IUNU
Totamot Imaging System is designed to train artificial intelligence to recognize crop risks under various conditions for tomatoes. It enables our IUNU horticulturalists to flag problems when crop risks are found. It allows us to provide actionable insights, including crop counts, issue detection, and crop steering, to growers to forecast the tomatoes better and increase profits.
Problem
Tomato growers want a solution that helps them to forecast tomatoes better and detect issues faster to make informed decisions.
Discovering the opportunity
Tomato plants can grow up to six to eight feet tall, depending on the varieties of tomatoes. Our team saw the opportunity to utilize our specialization imaging and analysis system to train artificial intelligence to learn the different crop risks for tomatoes and collect data. We also foresee that the data we collected from tomatoes can be used for other vine crops.
Designing for the Tomato Imaging System
The cameras on the trolley capture images at a fast speed. Thus, logging a crop risk needs to be quick and efficient. I collaborated with a computer vision engineer to understand how my work is being used on the backend and with internal horticulturalists to learn the current process of growing and forecasting tomatoes. This information is essential to ideate what the experience could look like.
Onboard Horticulturalists
After the user successfully logs into their account, they will go through a product tour. Although I was debating if users must sign up for an account to use Totamo Imaging System, I want to ensure our imagery annotators have a reference regarding who flagged the problem so they can communicate with them if further discussion is needed, as well as allow the users to track the issues they flagged.
The application is only available for internal use and is easy to use and self-explanatory. However, I didn’t want to assume that no users would experience any level of the learning curve, and we are also thinking of making this available to greenhouses later in the future. Thus, I believe a product tour must be included in the experience.
First Attemp
I categorized the risk into four categories: 1) Pests. 2) Diseases. 3) Environmental factors. 4) Need help to identify crop risks.
After learning the common crop risks associated with tomatoes with an internal horticulturist, I proposed flagging an issue with a risk category first because most licensed horticulturalists can quickly diagnose a crop risk at first glance. Even so, I also want to include a use case when it is hard to diagnose a crop risk due to many unknown factors. Thus, I added the Not Sure category to allow users to flag an issue by symptoms and add additional comments if needed.
I used images to represent each crop risk for the buttons so that users can look at the images as a reference to diagnose a crop risk accurately. Since there are typical problems tomatoes tend to encounter, I was aware that a horticulturist might create the same flag for the next tomato plant. Therefore, I added a behavior for the system to save the previous selection, but users can always change the details of the flag. Lastly, I designed the layout and the interface based on the greenhouse layout to allow users to visualize and create a flag more efficiently. For example, the last question asks the user which tomato plant this flag is associated with. The selection buttons are designed to look like the lanes at the greenhouse, so users only need to indicate if it is the tomato plant on their right or left has a problem. Each lane section has a geolocation tag that the trolley will recognize and capture images as the user creates flags. Hence, users don’t need to provide detailed location information to the flag.
Initially, I designed the Recent tab as a page for users to track all the flags they created. However, the CEO felt the need to use it for flag duplications to solve the need to create the same flag. I predicted this would create cognitive overload due to environmental factors and looking for the same flag by going through a long list.