Automatic watering of plants in a pot using plant recognition with CNN

The plant's growth is dependent on the soil moisture, but the required soil moisture threshold level is not the same for all the plant species. In the modern era with the combination of advanced Embedded and Computer Vision technologies, it is possible to detect the plant species on an Embedded platform and watering automatically. This paper discusses the implementation and working model which can detect the plant in a pot and water the plant automatically based on the threshold of the soil moisture and plant species. The complete model is built on Jetson Nano Nvidia GPU and a PCB which is de-signed according to the requirements, to control the motor pump and read sensor values. Python is used for the implementation of API in more customized way. The plant species and threshold values can be customized based on the climatic condition of any country. As the model is implemented for automatic watering of the plant in a pot, the calibration of the pot width is also done using the MobileNet SSD V2 algorithm to suggest the required soil moisture content and the water level in the tank.