[o]FlowerDex

Tools that went into FlowerDex

Inspiration

According to a study by the National Centers for Environmental Information, "The global temperature for January 2022 was the sixth highest for January in the 143-year NOAA record, which dates back to 1880. According to NCEI’s Global Annual Temperature Rankings Outlook, it is virtually certain (> 99.0%) that the year 2022 will rank among the 10 warmest years on record. " We are all witnessing global warming due and it has had a drastic effect on our plant and wildlife.

It is hard to deny our world's increasing dependency on technology. However, we believe that although it is becoming easier to live within the walls of our home, thanks to many innovative applications. We should enjoy the Earth we have today and continue aiding in our world's attempts on slowing down climate change in any way we can.

We wanted to create a fun way for people to get excited about going outside and exploring the plant life in order to inspire sustainability. We achieved this with FlowerDex - inspiring exploration and sustainability.

What it does

FlowerDex is a fun way to interact with the flowers that grow around us. We wanted to give our users the ability to take a picture of a flower they have and add it to their own FlowerDex, inspired by pokedex. Our image classification model was built in efforts of minimizing user input and putting more work on the computer, thanks technology! Here's how it works. Login or Signup for our app and you will be recieve a fresh FlowerDex to begin exploring and learning. In the FlowerDex page you will see it is empty! Return to the home page and snap a pic of a rose! Once classified you can go back the you FlowerDex page and see a new basge appear. You've officially got your starter flower! Clicking on the flower badge provides in depth information regarding the flower. Now that you have a general understanding for admire em' all!

How we built it

FlowerDex uses tenserflow and keras to build a transfer learned model using the mobilenet_v2 module giving our FlowerDex around 88% accuracy. For our frontend we created a React web app that ties into our python backend. We utilize two databases for all our user info, nosql and relational, hosted on cloud providers one of which is powered by Google Cloud.

Challenges we ran into

Our team faced many different challenges through the hackathon ranging from UI bugs with the frontend stack to database issues with backend stack. Our main challenge was learning how to implement machine learning within our application. No one on our team had experience with machine learning so it was the most difficult implementation within our project but also the most rewarding as we got it working correctly. Another main challenge we faced was allowing the data from the SQL database to interact with the flower data on React. Through hours of watching YouTube tutorials and reading GeeksForGeeks articles, we were able to overcome all of these challenges to reach our end goal.

Accomplishments that we're proud of

We are proud to have approximately 87% accuracy with our tensorflow model. It was each teammembers first time working with tenserflow and keras so it was quite the accomplishment for our team to have a working model that we were able to implement into our web app. We feel accomplished to have built an idea within 24 hours and are proud to present FlowerDex @ Citrus Hack 2022.

What we learned

Tenserflow, Keras, Transfer Learned Model

What's next for FlowerDex

Overall, our future goals for FlowerDex are to enhance it's frontend functionality as well as the TensorFlow model. Improving the frontend functionality will improve the user experience when using our application. This would mainly be UI changes through the app based on user experience research. Enhancing the TensorFlow model would improve the accuracy of the scanning function within our app. Our team also plans on integrating FlowerDex as a mobile application as well using React Native.