Leaf Dieses Prediction using ML

ABSTRACT :
The rice leaf suffers from several bacterial, viral, or fungaldiseases and these diseases reduce rice
production significantly. To sus-tain rice demand for a vast population globally, the recognition
of rice leafdiseases is crucially important. However, recognition of rice leaf diseaseis limited to
the image backgrounds and image capture conditions. Theconvolutional neural network (CNN)
based model is a hot research topicin the field of rice leaf disease recognition. But the existing
CNN-basedmodels drop in recognition rates severely on independent dataset and arelimited to
the learning of large scale network parameters. In this paper,we propose a novel CNN-based
model to recognize rice leaf diseases byreducing the network parameters. Using a novel dataset
of 4199 rice leafdisease images, a number of CNN-based models are trained to identifyfive
common rice leaf diseases. The proposed model achieves the highesttraining accuracy of 99.78%
and validation accuracy of 97.35%. The effec-tiveness of the proposed model is evaluated on a
set of independent riceleaf disease images with the best accuracy of 97.82% with an area
undercurve (AUC) of 0.99. Besides that, binary classification experiments havebeen carried out
and our proposed model achieves recognition rates of97%, 96%, 96%, 93%, and 95% for Blast,
Brownspot, Bacterial LeafBlight, Sheath Blight and Tungro, respectively. These results demon-
strate the effectiveness and superiority of our approach in comparison tothe state-of-the-art
CNN-based rice leaf disease recognition models.

EXISTING SYSTEM :
There exists many works for identifying and classifying the plant diseases specif-ically for
recognizing the rice leaf diseases [23]. In [2], a CNN-based frameworkis developed to identify
three different rice diseases and healthy images, while in[10] a CNN-based model is used to
detect various plant leaf diseases. In [6], prin-cipal component analysis (PCA) is applied to
remove the redundant informationand provide a vector with a reduced dimension for each rice
leaf disease image.Moreover, various classifiers were used to evaluate the performances and
SVMis found to achieve the best recognition rate of rice leaf diseases. In [29], diag-nosis of plant
disease based on CNN is achieved by extracting learned featuresvia perturbation, gradient and
reference based visualization using InceptionV3.Mixed layer is used to generate deep features
like shape, diversity of colors etc.It can remove 75% of parameters without affecting the
accuracy and loss. In[16], it is shown that CNN-based model is more effective than traditional
featureextractors including LBPH and HaarWT for recognizing rice blast disease.

DISADVANTAGES OF EXISTING SYSTEM :

1) Less accuracy

2)low Efficiency
PROPOSED SYSTEM :

This section describes our proposed method of recognizing rice leaf diseases. The
entire process is partitioned into different stages: beginning with the prepa- ration
of a novel training dataset, development of a novel CNN model, deep feature
extraction for training the model and finally, classification of the rice leaf diseases.

ADVANTAGES OF PROPOSED SYSTEM :

1) High accuracy

2)High efficiency

SYSTEM ARCHITECTURE :

HARDWARE & SOFTWARE REQUIREMENTS:

HARD REQUIRMENTS :

 System    :  i3 or above.
 Ram    :  4 GB.
 Hard Disk : 40 GB

SOFTWARE REQUIRMENTS :

 Operating system   : Windows8 or Above.

 Coding Language  : python