TY - GEN
T1 - Rice classification using scale conjugate gradient (SCG) backpropagation model and inception V3 model
AU - Parveen, Zahida
AU - Hasan, Yumnah
AU - Alam, Anzar
AU - Abbas, Hafsa
AU - Arif, Muhammad Umair
N1 - Publisher Copyright:
© Springer Nature Switzerland AG 2019.
PY - 2019
Y1 - 2019
N2 - Rice is one of the most consumed food crops all over the world. The classification of rice is a very crucial step after its cultivation. The identification of quality and class of rice is of great interest for researchers in this modern era. Although rice has many categories based on different features and characteristics like length, width, chalky and thickness, this particular research is based on the classification of different types of rice found in Pakistan. In this paper a comparative analysis of rice classification techniques of Neural Network (NN), which is Scale Conjugate Gradient Backpropagation Method (SCG) and Deep Neural Network (DNN) based on InceptionV3 model, is implemented on two variant types of datasets having single and collective samples of rice. There are nine different classes of rice present in each dataset which include Tota bland rice, Thalia 1121, Super Punjab, Kernel 1121, Steam 86, Basmati, super, Steam 85 and Super Tota. Each class contains 120 samples for training. Total 18 samples of rice are used for testing the network accuracy of each dataset. In this research, data is classified based on maximum area of the grain. The results reveal that InceptionV3 model has better accuracy as compared to SCG method. However, some false classification has also occurred due to the similar readings of area, less difference in the values of extracted features, similar structure and low level noise.
AB - Rice is one of the most consumed food crops all over the world. The classification of rice is a very crucial step after its cultivation. The identification of quality and class of rice is of great interest for researchers in this modern era. Although rice has many categories based on different features and characteristics like length, width, chalky and thickness, this particular research is based on the classification of different types of rice found in Pakistan. In this paper a comparative analysis of rice classification techniques of Neural Network (NN), which is Scale Conjugate Gradient Backpropagation Method (SCG) and Deep Neural Network (DNN) based on InceptionV3 model, is implemented on two variant types of datasets having single and collective samples of rice. There are nine different classes of rice present in each dataset which include Tota bland rice, Thalia 1121, Super Punjab, Kernel 1121, Steam 86, Basmati, super, Steam 85 and Super Tota. Each class contains 120 samples for training. Total 18 samples of rice are used for testing the network accuracy of each dataset. In this research, data is classified based on maximum area of the grain. The results reveal that InceptionV3 model has better accuracy as compared to SCG method. However, some false classification has also occurred due to the similar readings of area, less difference in the values of extracted features, similar structure and low level noise.
KW - Backpropagation
KW - Deep neural network
KW - Inception V3
KW - Scale conjugate gradient
KW - TensorFlow
UR - https://www.scopus.com/pages/publications/85057082916
U2 - 10.1007/978-3-030-01177-2_10
DO - 10.1007/978-3-030-01177-2_10
M3 - Conference contribution
AN - SCOPUS:85057082916
SN - 9783030011765
T3 - Advances in Intelligent Systems and Computing
SP - 129
EP - 141
BT - Intelligent Computing - Proceedings of the 2018 Computing Conference
A2 - Kapoor, Supriya
A2 - Bhatia, Rahul
A2 - Arai, Kohei
PB - Springer Verlag
T2 - Computing Conference, 2018
Y2 - 10 July 2018 through 12 July 2018
ER -