The 4th International Conference on Next Generation Computing 2018 Automatic Disease Detection in Wheat Crop using Convolution Neural Network Altaf Hussain Department of Computer Science, Islamia Collage University Peshawar, Peshawar, Pakistan develper.altaf@gmail.com Mohsin Ahmad Department of Computer Science, Islamia Collage University Peshawar, Peshawar, Pakistan mosinshmad938@yahoo.com Imran Ahmad Mughal Department of Computer Science, Islamia Collage University Peshawar, Peshawar, Pakistan imranmx@yahoo.com Haider Ali Department of Computer Science, Islamia Collage University Peshawar, Peshawar, Pakistan haider6233icp@gmail.com Abstract―Crop’s diseases reduce the production and mainly responsible for the economic losses in agricultural industry, worldwide. For the betterment of human health, the diseases in crops must be control and effectively monitor. Previously, researchers have used hand-crafted- features for image classification and recognition. Nowadays, the development in Deep Learning has allowed researchers to drastically improve the accuracy of object detection and classification. In this paper, we used a deep-learning framework to classify wheat diseases using images captured in-place by camera devices, with various resolutions. Our dataset contains four categories of wheat disease like stem rust, yellow rust, Powderly and normal. Each category contained 2,207 images. To train our classifier we have used the Convolutional Neural Network (CNN). One of the biggest advantage of CNN is to automatically extract features by processing the raw images directly. Our model obtained 84.54% accuracy and it can be used as a practical tool for farmers to protect wheat crop, against aforementioned diseases. Keywords― Image Classification, Convolutional neural network, AlexNet architecture, wheat disease I. Introduction Plants can be seen all over in our surrounding, accordingly a broad variety of plants on earth. For example, more than 240,500 species of plants have been named and recorded statistically the role of plants is very important as it improves the climate, serve soil, and water. Especially plants provide us the food and oxygen. In fact, every plant has its own characteristics for example economics value, habits and other morphology. For that reason, Farmers have large range of diversity for selecting various suitable crops and finding the suitable pesticides for plant. Wheat is a kind of plant grown throughout the world for its majorly nutritious and usable grain. It is one of the top most produced crop in the world, along with rice and corn [1].Cultivation of wheat over 6,000 years and likely originates in the fertile field, along with other stable crops. Disease on wheat leads to the important reduction in the quality and quantity of the product of agricultural [2]. In wheat most common disease are stem rust, yellow rust and Powderly. To understand effectively the disease of plant leaf experiments prefer visually observable pattern of wheat plant disease [2]. Monitoring of leaf disease play important role for the successful cultivation of crop in the farm. In the early days, expert person manually monitors the wheat disease. To achieve this, the experts must have agriculture train and deep knowledge of various field as well as experience in the diseases symptoms and also have knowledge of the causes of diseases. In some advance countries, farmers may have to go long distances to contact experts, these form consulting experts is very expensive and time consuming. Automatic detection of wheat diseases is a beneficial research as it validate to help in monitoring large fields of crops, and automatically detect the symptoms of diseases immediately as they appear on wheat leaves[3]. Protecting plants from diseases is crucial by improving the quantity and the quality of the crops. Thus, providing primary detection and picking out the target disease is more helpful by choosing correct treatment and stopping the disease from spreading [4-6]. Deep learning is a new trend in machine learning and it achieves helpful results in many research fields, such as computer vision, drug design and bioinformatics[4] etc. The advantage of deep learning is the ability to exploit directly raw data without using the handcrafted features[4, 6]. Large amounts of data are generated every day. Hence, these data can be used in order to train a deep learning model. Secondly, the power of computing provided by GPU (Graphics Processing Unit) and HPC (High Performance Computing) make possible the training of deep models and leveraging the parallelism of computing. The aim of the present study is to introduce deep learning as an approach for classifying plant diseases, focusing on images of leaves. To improve classification accuracy, we used the Convolutional neural network as sajjad et al. [7] used SVM for classification. Besides, deep learning models have the ability to use raw data directly without feature engineering. Moreover, deep learning models offer the possibility of transfer learning from another task by using already trained models on large datasets. In the present paper, we suggest the use of deep learning and specifically