แอปพลิเคชันจำแนกและสืบค้นข้อมูลกระสุนปืนจากภาพปลอกกระสุนปืน

• Main Authors: ณภัทร พิพัฒน์โรจนกมล, นนท์ ศรีคำ
• Other Authors: รัชลิดา ลิปิกรณ์
• Format: Senior Project
• Language: Thai
• Published: จุฬาลงกรณ์มหาวิทยาลัย 2019
• Subjects: กระสุนปืน > ฐานข้อมูล; โปรแกรมประยุกต์; ระบบการจัดเก็บและค้นคืนสารสนเทศ > กระสุนปืน; Bullets > Databases; Application software; Information storage and retrieval systems > Bullets
• Online Access: https://digiverse.chula.ac.th/Info/item/dc:10399
• Institution: Chulalongkorn University

The proposed project entitled "Application for bullet classification and information retrieval from a bullet shell image" was developed to help a user to easily identify and search for the details of a bullet, such as the manufacturing country. Moreover, the application can also reduce the chance of human errors. The application was designed and developed for bullet classification from the features of the logo and bullet shell, and it returns the manufacturing country as a result. The application allows a user to select an image file and convolutional neural network is used to classify a bullet. Two convolutional neural networks called AlexNet and VGGNet are used in this project. The convolutional network was trained using an image data set containing 687 images of size 227x227 pixels from 14 countries. The test data were divided into two sets. The first set contains test data that were randomly selected from the training set and the second set contains unknown bullets. From the experimental results, the application can correctly classify 60% of the first test set and can correctly classify about 45% of the second test set.