SCIENCE AND TECHNOLOGY NAKHON SAWAN RAJABHAT UNIVERSITY JOURNAL

การสร้างคำบรรยายภาพด้วยแบบจำลอง CLIP Prefix Caption บนชุดข้อมูล Traffy Fondue

วสิศ ลิ้มประเสริฐ — Fri, 08 Dec 2023 00:00:00 +0700

Traffy Fondue is a grievance system provided by the Bangkok Metropolitan Administration to receive opinions and suggestions that citizens have on the city. However, due to the amount of information from many users, there is still unclear reporting, such as descriptions and images are inconsistent, which makes it difficult for the receiving officer to coordinate and solve the problem. Therefore, our team proposed a data clustering method to increase the ability of clustering data to be more convenient by using data processing techniques. In this research, the presenter applied the CLIP Prefix Caption model for creating captions and allowing the system to group words or search for related problems. The CLIP, CLIP Prefix Caption and GPT-2 models were created captions using images from Traffy Fondue. The experimental results can be summarized as follows: BLEU 0.93% and ROUGE-1 16.39%. The result is not good enough for real applications. The experiment was further reorganized by proposing to group images using vectors from Prefix Embeddings instead of captioning directly from the image. The results indicate embedding can be applied for further development.

การวิเคราะห์ทัศนคติเกี่ยวกับการจัดการภัยพิบัติในประเทศไทยบนทวิตเตอร์

ปกป้อง ส่องเมือง — Fri, 08 Dec 2023 00:00:00 +0700

The purpose of this research was to study the needs and attitudes of Thai people towards disaster management in each situation of the Thai government. The top three hashtags on Twitter are a chemical leak, The King Kaew factory fire, and Flooding in Ubon. As a result, Sentiment Analysis knowledge was applied. The data obtained will be used with two data sets: all data sets and the data set that will be used to group the data (Clustering) to obtain a more specific data set that only conveys the user's attitude. The researchers created four models: Logistic Regression, Decision Tree, Random Forest and Support Vector Machine (SVM). To compare the two dataset’s sentiment analysis predictions. Clustering the messages first can help yield better results. However, the data is reduced in size as a result of segmentation. As a result, employing The Synthetic Minority Oversampling Technique (SMOTE) aids in the resolution of the problem. The best model across all hashtags is Decision Tree, which has an Accuracy Score of 93% in #Chemical Leaks and 91% in #KingkaewFire and #Flooding in Ubon. Word Cloud was used to study requirements or what the most people say, both positively and negatively. The words 'government' and 'politics' were discovered to be frequently used in every hashtag's positive and negative connotations. This study concludes that the Thai government's disaster management is insufficient to meet the needs of the Thai people because shows primarily on the negative.

การพัฒนาอุปกรณ์ตรวจหาและแจ้งเตือนการบุกรุกจากข้อมูลเสียงใต้น้ำ

ปริพรรห์ วิวัฒน์นิธิกร — Fri, 08 Dec 2023 00:00:00 +0700

This research aims to develop an intrusion detection and warning device based on underwater sound data. with a purpose 1) to develop a device for detecting intrusion into the area by underwater audio data 2) to develop a video recording and intrusion warning device using underwater audio data for conservation work and 3) to analyze Volume for building intrusion detection devices There are 5 steps in researching as follows: 1) installing a hydrophone underwater 2) getting sound data from a hydrophone 3) importing data into the noise filtering process 4) specifying the volume and 5) video recording and warning when the volume exceeds the specified level. This research demonstrated that the device can filter out the frequencies of the sound as a low pass filter. by making the mid-frequency, low-frequency and high-frequency bands can be output with less noise. while recording video when the sound exceeds the limit with notification via the LINE application. In addition, research on the development of intrusion detection and alarm devices showed that the frequency range of 0 to 1,000 Hz is more suitable for the device's performance than other frequency ranges.