Classification of potatoes into edible and non-edible by using deep learning technique

Singh, Nikhil Kumar

Classification of potatoes into edible and non-edible by using deep learning technique

Date

2022-08

Authors

Singh, Nikhil Kumar

Publisher

G.B. Pant University of Agriculture and Technology, Pantnagar, District Udham Singh Nagar, Uttarakhand. PIN - 263145

Abstract

Potato, one of starchy vegetables, needs to be classified before entering industry for assuring the product quality. However, it is time-consuming, tedious, labor-intensive, inaccurate and expensive to assess qualitatively and classified manually. In this study, we propose an efficient and effective machine vision system based on the state-of-the-art deep learning techniques and stacking ensemble technique to offer a non-destructive and cost-effective solution for classification of ‘potatoes’ freshness and appearance. It has been trained and tested the performance of various deep learning models including ResNet, MobileNet, InceptionV3 and VGG16 to find the best model for the classification of potatoes into edible and non-edible categories. The system was trained and tested on potatoes data sets which were collected at constant distance of 20 cm for image capturing. The accuracy of used models is found to be 89 %, 68%, 95%, and 99% respectively and the precision and recall of the proposed model was best for MobileNet with uniform internet speed for each model. The experimental results show that the accuracy, precision and recall achieves with limited sample easily which tells the overall performance of the proposed model. The proposed technique for classification has less parameters and lower computation complexity than popular networks. The result proves that it can be extended to other tasks about classification.