E-FARM - a Mobile Application for Predicting the Plant Diseases

EasyChair Preprint no. 8579

Abstract

As you all know agriculture is major backbone of our country. However, due to the lack of proper observation of each leaf by the farmer with his/her physical presence, every time farmer is ending up with heavy losses. So, it is essential to build an automated system which can detect the diseases by reducing the load on the farmer, and thus helping in easy and efficient yield. So, in this paper we develop an automated system in detecting the plant leaf diseases using various machine learning (ML) algorithms such as K-nearest neighbors (KNN), support-vector machine (SVM), convolutional neural network (CNN) and visual geometry group (VGG)16. The plant leaf diseases dataset was called from Kaggle website with 2411 images of size 256*256.Later the images were denoised, where the denoising effects are resizing and removing the blur images. Furthermore, features are extraction by edge detection using Gabor filter. The dataset is divided into training and testing of 80-20 and 70-30 % respectively. Moreover, we also evaluated the performance of the above ML algorithms in-terms of accuracy, precision, recall and F1 score. From the results, it is observed that for a of 80-20% an accuracy of 39% for KNN, 76% for SVM, 98% for CNN model and 99% for VGG16 at the 10th epoch. With 70-30%, accuracy of 35% for KNN, 65% for SVM, 94.5% for CNN model and 96% for VGG 16 is observed. In both the cases, VGG16 has got highest accuracy at the 10th epoch when evaluated with confusion matrix. Finally, we also developed a mobile application by implementing VGG16 inside it, for the farmers to check the status of leaves on the go.

Keyphrases: feature extraction, machine learning, mobile application, object detection, Plant Disease

@Booklet{EasyChair:8579,
  author = {Anita Patil and K.Viswavardhan Reddy},
  title = {E-FARM - a Mobile Application for Predicting the Plant Diseases},
  howpublished = {EasyChair Preprint no. 8579},

  year = {EasyChair, 2022}}