แจ้งเอกสารไม่ครบถ้วน, ไม่ตรงกับชื่อเรื่อง หรือมีข้อผิดพลาดเกี่ยวกับเอกสาร ติดต่อที่นี่ ==>
หากไม่มีอีเมลผู้รับให้กรอก thailis-noc@uni.net.th
หากไม่มีอีเมลผู้รับให้กรอก thailis-noc@uni.net.th
Niracha Chaiwong. Classifying saleability of lime by using convolutional neural network (CNN) approach. Master's Degree(Logistics and Supply Chain Systems Engineering). Thammasat University. Thammasat University Library. : Thammasat University, 2021.
Classifying saleability of lime by using convolutional neural network (CNN) approach
Name: Niracha Chaiwong
keyword:
Neural network
Abstract:
Nowadays, there are many applications that have been used in agriculture field to facilitate human workers. Many researches have been focused on fruits grading based on appearance features. Since lime is one of Thai economic crop that have high demand and price all the year, a lot of limes are continually exported into the market. The appearance of limes strongly impact the market value. So, the quality inspection is needed to avoid selling bad limes to customers which cause greatly affects to the customer base. Determining saleability of lime can be done by using humans or machines. Using human workers is time-consuming, high cost and not accurate. Although there are fruit grading machines, they are expensive and difficult to reach for general farmers. This research has proposed method to classify saleability of lime into three categories which are buy, maybe, and not buy by using Convolutional Neural Network (CNN) because it is a deep neural network that has high performance in terms of images classification tasks. Hyperparameters tuning process has been used to search for the best model performance by using full factorial design. The hyperparameters that have been considered in the model are Epoch, Learning Rate, Decaying Rate and Momentum. Two experiments of hyperparameters tuning were constructed. The first experiment constructed by initial 2^5 full factorial by setting low and high value. After that, the result from first experiment was analysed by using Minitab to remove unimportant factors. The model performance in this research has been evaluated by F1-score which the first experiment gave the highest results as 76%, 79%, 80% in training model, 73%, 77%, 77% in validation and 74%, 77%, 77% of testing for buy, maybe and not buy respectively
Thammasat University. Thammasat University Library
Address:
BANGKOK
Email:
preserv@tu.ac.th
Name: Warut Pannakkong
Role:
advisor
Role:
co-advisor
Created:
2021
Modified:
2023-04-20
Issued:
2023-04-20
วิทยานิพนธ์/Thesis
application/pdf
eng
DegreeName: Master of Engineering
Level: Master's Degree
Descipline: Logistics and Supply Chain Systems Engineering
Grantor: Thammasat University
©copyrights Thammasat University
RightsAccess:
ใช้เวลา
0.025009 วินาที
Niracha Chaiwong
| Title | Contributor | Type |
|---|---|---|
| Classifying saleability of lime by using convolutional neural network (CNN) approach
มหาวิทยาลัยธรรมศาสตร์ Niracha Chaiwong | Warut Pannakkong Jirachai Buddhakulsomsiri | วิทยานิพนธ์/Thesis |
Warut Pannakkong
Jirachai Buddhakulsomsiri