|
Apply of Deep Learning Techniques to Measure the Sweetness Level of Watermelon via Smartphone |
|---|---|
| รหัสดีโอไอ | |
| Creator | Nattavadee Hongboonmee |
| Title | Apply of Deep Learning Techniques to Measure the Sweetness Level of Watermelon via Smartphone |
| Contributor | Nutthapong Jantawong |
| Publisher | Faculty of Information Science and Technology, Mahanakorn University of Technology |
| Publication Year | 2563 |
| Journal Title | Journal of Information Science and Technology |
| Journal Vol. | 10 |
| Journal No. | 1 |
| Page no. | 59-69 |
| Keyword | Watermelon Sweetness, Image Classification, Deep Learning, Convolution Neural-Network, Smartphone |
| URL Website | https://tci-thaijo.org/index.php/JIST |
| Website title | Journal of Information Science and Technology |
| ISSN | 2651-1053 |
| Abstract | This research proposed the development of automated system for analyzingsweetness and watermelon varieties from photos using deep learning techniques for use onsmartphones. To help the public who would like to know the names of varieties and sweetness ofwatermelons. The main components of the system include: (1) Modeling, classifying varietiesand sweetness levels of watermelons with deep learning neural network through the TensorFlowlibrary, using the InceptionV3 and MobileNet algorithms to compare image classification. Inwhich the trainers are able to classify 4 types of images, each type of 100 images, and trainingour model for 500 rounds. The result shows that the model from the InceptionV3 algorithm hasthe same accuracy as the model from the MobileNet algorithm. The accuracy is 9 7 .2 0 % .Therefore considering the model size obtained from learning, it found that MobileNet modelsize is smaller than InceptionV3, so choose the model from MobileNet to develop the systemfurther. (2) Using the model from MobileNet algorithm to develop application on smartphones,which developed by Android Studio program. Results of the user satisfaction test, found thatthe average satisfaction is 4.34 , standard deviation 0.62 , it at good level. In conclusion, thisapplication is effective and can use in real life. |
