Ruohan Zhang¹, Yaotian Zhang1, Yu Sun², ¹Fairmont Preparatory Academy, USA, ²California State Polytechnic University, USA

All living things including plants need water to survive, and agriculture is the world’s biggest user of water [4]. Unfortunately, in a worst-case scenario, over-watering and drying up cause both water waste and the plant’s death [5]. Guided by this problem that is frequently occurring around the world, we designed an app to determine if the plant needs to be watered or not by capturing pictures of a certain plant and training an AI to compare whether the soil in the pot is dry or wet. In this program, we use Raspberry Pi to capture an image of the plant every 10 seconds,in which the Python code using TensorFlow inside the Raspberry Pi will determine the moisture level of the soil [6][7]. The result will be posted to Firebase with a timestamp, and lastly, we have a mobile app that can display the result from Firebase to the user. We published our application to Apple’s App Store and the Google Play Store, and public installation of the app means that it can have more widespread usage. An experiment was performed to determine whether the application’s model can accurately determine soil moisture [8]. The results indicate that the model is very accurate for the vast majority of soil samples under various lighting conditions.

Python, Flutter, Machine Learning, Firebase