UM Plantify is a proof of concept plant identification app.
UM Plantify is a proof-of-concept mobile application for plant identification for the Nichols Arboretum at the University of Michigan. It runs locally on a computer that allows a user to use the application through the Expo app. This application can identify 20 species from a photo with a training accuracy of over 96% and an average testing accuracy of 90%. After predicting the species, the app displays a photo and information about the predicted species. In addition, we included the ability to browse our database of supported plant species as well as a map of the Nichols Arboretum.

Project Overview

ProjectOverview:
University of Michigan - Information Technology Services
Summer Internship 2020
‍
Project Members: Sydney Bruce, Camille Lewis, Randy Andrews, Alexis Ashby, Vincent Cao, Taylor Gribble
‍Role: User Experience Designer

Timeline: 10 Weeks

The Process

As a group we were given the freedom to decide how we wanted to go about completing our project in the allotted time. After selecting our scope for the project, which was the creation of a plant identification application for the Nichols Arboretum, we as a group organized ourselves into teams. These teams were created based off skill set as well as personal preferences.

Comparative Analysis

As s member of the design team we had to come up with a plan for how we wanted to begin the design process. We knew other apps and websites like this existed and felt a comparative analysis would be the best way to gather design and functionality ideas. We focused mainly direct and secondary applications that had basic photo scanning and identification functionality. We narrowed our analysis down to three applications:

Overall Observations:‍

Key Takeaways:

Target Audience / Inclusivity

Taking into consideration the affiliation the app would have with the University we wanted to create a general list of potential users to take into consideration when designing. We separated them into Public Users, Researchers, and University Related Users.

With these various users in mind we also created a list of potential elements to include to allow for inclusivity for the wide range of potential users. We especially took into account the large range in technical knowledge of potential users to make sure that our app could be used by users of all ages. Below are characteristics we felt were most important to consider moving forward to maximize inclusivity:

Functionality

Top Priority:

If we have time:

Future Expansions:

Design Elements

Branding

Since the application is affiliated with the University of Michigan I had to abide by the university's branding guidelines when designing, especially in relation to the logo.

Sketches

Low-Fi Prototype

Design Iterations

After creating a basic layout and flow of the application, we went back in and created various iterations of the home screen that we felt would best layout the three primary functions of the app. Below are our three basic layouts.
Home Page:
We also iterated upon final design after deciding on which home screen layout we felt was the best. We added potential color scheme as well as final flow of the application.
Final Layout:

Logo Iterations

Final Logo Design

Style Guide

We aimed to select a style that would reflect the outdoors and give the app a natural look. Below is the color palette and typography. We chose a more blue-green color palette rather than a strictly green color palette to give it neutrality. The darker blue was our primary color (#042940) while the middle blue became our accent color (#588C7E). We felt it matched the natural colors we were aiming, giving it a natural feel.

Mobile Application

We were asked to present our final product at the end of year showcase in front of ITS and University of Michigan Staff. The app is a proof of concept and can run off local devices when given access through the Expo app. We have detailed instructions for both how to run the app locally as well as how others who wish to add to this application can both attain and edit the existing app.
Welcome to the Nichols Arboretum! It’s a beautiful day, and we’re glad you could join us. Enjoy the scenery and Michigan plant life as you walk down one of the numerous trails. I see you stumbled upon an interesting plant! Are you wondering what it is? We have an app that can help you with that!

UM Plantify is a simple, easy-to-use mobile app for plant identification. Its three main functions are clearly displayed on the home page. The help button in the upper-right corner takes you to the Help page that has instructions for taking a good photo as well as descriptions of the app features.

Choosing “Take Photo” brings up your camera and allows you to take a photo of the plant you’d like to identify. After taking a photo, you’re then prompted to confirm the photo in case you’d like to retake it. After choosing “Yes”, the image is processed by the machine learning model, and its prediction is displayed along with information about the species. The displayed information includes a photo of the plant along with helpful information that you can also find in the Database.

The Database feature allows you to scroll through the species that the app supports. Choosing a plant, such as Scarlet Beebalm, in the database brings you to the information page for that plant. The displayed information includes a description of its appearance, information about the plants origins and locations, and a fun fact. Use this database to learn about any of the supported plants without having to take a photo of it.

If you’re using this app while you’re in the Arb, then we included the Map feature to see your live location in the Arboretum. Looking at the screen you can see the location services at work showing the user where they are within the arb. Users also have the option to view the labeled Arboretum map. This allows users to view the static map in comparison to the location services to see what landmarks and trails are in their surrounding area.

Reflection and Next Steps:

We created a simple app that serves as a great proof-of-concept for plant identification. Despite the fact that our app currently identifies 20 species, we have set up an easy process to add more species to the classification model, it’s just a matter of obtaining more photos and re-running the script to train the model. Currently the app's accuracy is over 90%, but it has had trouble getting the correct classification for spider flowers. This can be fixed by adding more data (pictures) of spider flowers to the model. We hope to offer more features such as image recognition, photo tracking on a map, and User profiles. We hope that this app will serve as a strong foundation for a tool that allows anyone to learn more about plant species and identification and we hope for it to be picked up and further integrated in the future.
Ideas for improvements to the app:

UM Plantify

Plant Identification App

ReadMe

Read, explore, and share your favorite books.

Roomiez

Find the Perfect Roommate.

©2020 Sydney Bruce. All Rights Reserved