Identification of Edible and Non-Edible Mushroom Through Convolution Neural Network

Abstract

Mushroom is one among the most popular consumed food in India. In India people are cultivating mushroom as viable income source for their livelihood. Now-a-days deep learning is being applied to process big data and vision related applications. Recent smart devices can be utilized for automated edibility diagnosis of mushroom using deep convolution neural network (CNN) it has revealed a remarkable performance capability in all its sphere of research activities. DCNN works on static dataset. The models on which it applies will pose as well determine its requirement for training. This paper presents a classification tool for edibility detection of mushroom through deep CNN. Better performance is obtained by tuning the hyper-parameters and through adjustments in pooling combinations in order to obtain real time inference suitably. DCNN has been trained with a data set of segmentation as train and test sets. Performance is analyzed on sNet, Lenet, AlxNet, cNET network architectures. DCNN results are comparatively better in its performance.