Honey Bee Bearing Pollen and Non-Pollen Image Classification, VGG16 Transfer Learning Method using Different Optimizing Functions
Pankaj Kumar Kandpal1, Ashish Mehta2, Avinash Sharma3
1Pankaj Kumar Kandpal*, Department of Computer Science, D. S. B Campus, Kumaun University, Nainital, Uttarakhand, India.
2Ashish Mehta, Department of Computer Science, D. S. B Campus, Kumaun University, Nainital, Uttarakhand, India.
3Avinash Sharma, Deprtment of Computer Science and Engineering, Maha Markandeshwer University, Mullana, Ambala, Haryana, India.
Manuscript received on November 17, 2019. | Revised Manuscript received on 28 November, 2019. | Manuscript published on December 10, 2019. | PP: 3655-3659 | Volume-9 Issue-2, December 2019. | Retrieval Number: B8130129219/2019©BEIESP | DOI: 10.35940/ijitee.B8130.129219
Open Access | Ethics and Policies | Cite | Mendeley | Indexing and Abstracting
© The Authors. Blue Eyes Intelligence Engineering and Sciences Publication (BEIESP). This is an open access article under the CC-BY-NC-ND license (http://creativecommons.org/licenses/by-nc-nd/4.0/)
Abstract: The deep leaning has been the prominent and interesting field for researchers for last few years. There are several studies are being done on real problems in several streams of Science. In the deep leaning there are huge resources and labeled dataset are required to train the model. In convolutional neural network (CNN) training with random weights take a considerable amount of time to optimize the weights. To minimize efforts and other resources and deal with large, challenging dataset; which generally very complexed to use in the conventional training methods. New concept of deep learning has been evolved and emerge as prominent methods called transfer learning in last few years. In the transfer leaning, pre-trained model VGG16, which had been trained using 1000 different classes of Images. The VGG16 architecture having 16 CCN layers with ReLu activation function. In this study we took the Honey Bee image dataset consists of two type of the images of honey bee, some images are bearing pollen honey bees and remaining are non-bearing pollen honey bees. We got learn the model with dataset and classified problems using three optimization functions: Adaptive Moment Estimation Algorithm (ADAM), Root Mean Square Propagation (RMSprop) and Stochastic gradient descent (SGD). we compared the results, and found that Adaptive Moment Estimation (ADAM) function performed best, and got 91.94 percent Accuracy and 0.2027 losses.
Keywords: Deep Learning; Transfer learning; Honey Bee Bearing Pollen; Accuracy, Classification.
Scope of the Article: Classification