Article Preview
Top1. Introduction
A large majority of the population in India depends on agricultural production. India ranks second in the agriculture output worldwide. In India, farmers cultivate a great diversity of crops. The food welfare and plants full of life and lively are interpersonally linked. The(FAO) displays that illness among plants and pests and diseases paves the way to become the cause of food production loss worldwide, approximately 20% - 40%, which is a big threat to food security. The use of pests is a way to protect crops to this law makes way for danger in the security of food, but applying them on crops leads to a negative impact on biodiversity and a to solve the requirement of expanding population. The situation is further complicated as nowadays of danger to humanity intense and persistent. Diseases can spread freely than in old times. The symptom of infected plants disease become visible superficially on leaves there become visible resulted in rotten leaves or rotten fruits. The disease was created due to pathogens such as Fungi, Bacteria, and Viruses. In Bhopal, the public observed that vegetables are being irrigated through sewage water flowing in a big drain in some areas. We can only imagine the impact of these vegetables appearing in our kitchens and resulting in health hazards. A report of FAQ says that there is a requirement to boost food production up to 65% to 70% by 2050 to fulfill the necessity of food in the world. 40% of food production is destroyed by infected plants' use of pests. In this category, Maharashtra ranks on top where farmers' suicide due to major reasons is the failure of crops.
There is a need for some method to detect diseases on the plant in the early stages. The identification and tabulation or classification of disease in the leaf is the key to prevent agriculture loss. For this purpose, Deep Learning is used. Deep Learning, in many cases, has been proved as a game-changer in solving complicated problems in a short span of time.
A variety of different methods and a lot of algorithms may be implemented for the mentioned purpose. For example, these include Linear Regression, Logistic Regression, K-Nearest Neighbours (KNN), Naive Bayes (NB), Gaussian Models, Decision Trees, +78+—Random Forest, Clustering and Support Vector Machines (SVM). In comparison to conventional methods, the above-said methods give precise forecasting helping in the flow of making a decision. Now in solving complicated problems, Deep Learning methods are being used and Capable to do pretty soon. (Anderla et al., 2019) Deep Learning is can also be useful for many other assignments as it is already a State-of-the-art methodology for the land cover classification task.
1.1 Deep Learning in Image Processing
Deep Learning is a subset of Machine Learning, which is one of the Machine Learning techniques. (Qiu et al., 2019) Deep Learning instructs computers to act in such a way as human beings perform inherently. Deep learning may be a great potential for generating a successful model, framework, and method for achieving sustainable computing. It is the main technology back automated cars, capable of determining a sign of stop or discrimination between road and pedestrians. In Deep Learning machine is being trained for assortment directly from images, sounds, or text. The DL models can acquire progressive validity. A NN contains several layers and categorized data and Neural Network architecture used to train a model.
A large amount of categorized data is needed in DL. It also essential extraordinary computing power. High Geared GPUs have a parallel architecture that is adequate for DL. Utmost DL methods used Neural Network architectures, so that is why Deep Learning is also termed as Deep Neural Networks. DNNs can be classified into three main types: convolutional neural networks (CNNs), deep belief networks (DBNs), and recurrent neural networks (RNNs) (Liao et al., 2019).The name “deep” commonly cite for NN hidden layers. Conventional NN contains 2-3 hidden layers, but DL can contain as 150. The most popular Deep Neural Network is a Convolutional Neural Network (CNN).
There are commonly two ways to process the image, i.e., using grayscale and using RGB values:
- 1.
Gray Scale: In this image is converted into white to black, which grayscale and a range of shades. Every pixel is attached to a value that confers to how dark it is.
- 2.
RGB Values: The colors are represented as RGB values, then each pixel extracted and set the result for the array.