Agriculture Automation using Deep Learning Methods Implemented using Keras
Machiraju Yashwanth, Muramalla Lahari Chandra, K V Pallavi, Danish Showkat, P. Satish Kumar
Abstract
The primitive way to do away with weeds was to remove them manually. As time went by, the farmers started using herbicides to kill the weeds. The excessive usage of weedicides can cause severe health problems to the agricultural workers/farmers, can also contaminate the soil and the water. New methods have to be introduced to reduce the usage of weedicide. Even after the advancement in research, a lot of readily available solutions are not being implemented at the grass-root level. Addressing the problem, a better solution has been proposed to minimize the usage of the herbicide by classifying the plant images into the weed and the crop for selective spraying of the herbicide. The first step towards it is to differentiate between the crops and the weed. Image Classification Technique has been implemented using the Deep Learning function. A maximum efficiency of 96.3% was achieved with just 250 images of each plant in the dataset. The proposed model may easily be dumped in the Raspberry Pi and the selective spraying may be performed with the help of an attached sprayer. This setup may be installed on a tractor or a drone for real-time implementation.