Crop image classification using convolutional neural network
M. M. Sardeshmukh, Midhun Chakkaravarthy, Sagar Shinde, Divya Midhun Chakkaravarthy
Abstract
A crop image classification using convolutional neural network is proposed in the paper. Classification of crop images is important and required in many applications such as yield prediction, decease detection etc. (Yang et al., 2020)(Kavitha et al., 2022). The main challenges are availability of the large dataset and extraction of meaningful features to describe a class of image(Barbedo, 2018). We have proposed a convolutional neural network to and the pre-trained models like VGG 16 and Resnet 50 for crop image classification. The pre-trained models trained on millions of images for a very large class size. The results shows that VGG 16 can be best used for our application as gives the accuracy of more than 98 %. The CNN training accuracy is 93 % but testing accuracy is only 42%. This is due to the lack of training data available. The accuracy of the CNN can be improved using large dataset. The Resnet 50 fails for crop image classification.