Pub Date : 2021-12-21DOI: 10.1109/NICS54270.2021.9701547
Sinh Van Nguyen, Dung Anh Nguyen, Lam-Son Pham
Digital transformation is one of the most popular keyword in recent years. It is not only a trend in science research based on the development of information technology, but also a proposed duty that applied in the companies or organizations nowadays. Digitalization of administrative documents is therefore considered as the first step in digital transformation of public organization. Through the digitizing process, the information that were in written format or hard copies will be converted into digital format (e.g. document files) to serve for storing, mining, processing and managing the documents. This paper presents a method to build a web application for digitizing the administrative documents applied in most public organizations. The method is based on the OCR (Optical Character Recognition) combined with the image processing techniques. Our digital process is implemented as following steps. (i) Scanning the hard copies of the administrative documents. (ii) Removing noise data and filtering necessary information in the content based on image processing technique. (iii) Classifying automatically the acquired contents into the respective components of a template form following the structured format of Vietnam Government. (iv) Generating automatically a document file. The application can process a document with a single or multiple pages. To compare with similar applications, our application is processed very fast, without limitation of pages for each document and obtained accuracy as our expectation.
{"title":"Digitalization of Administrative Documents A Digital Transformation Step in Practice","authors":"Sinh Van Nguyen, Dung Anh Nguyen, Lam-Son Pham","doi":"10.1109/NICS54270.2021.9701547","DOIUrl":"https://doi.org/10.1109/NICS54270.2021.9701547","url":null,"abstract":"Digital transformation is one of the most popular keyword in recent years. It is not only a trend in science research based on the development of information technology, but also a proposed duty that applied in the companies or organizations nowadays. Digitalization of administrative documents is therefore considered as the first step in digital transformation of public organization. Through the digitizing process, the information that were in written format or hard copies will be converted into digital format (e.g. document files) to serve for storing, mining, processing and managing the documents. This paper presents a method to build a web application for digitizing the administrative documents applied in most public organizations. The method is based on the OCR (Optical Character Recognition) combined with the image processing techniques. Our digital process is implemented as following steps. (i) Scanning the hard copies of the administrative documents. (ii) Removing noise data and filtering necessary information in the content based on image processing technique. (iii) Classifying automatically the acquired contents into the respective components of a template form following the structured format of Vietnam Government. (iv) Generating automatically a document file. The application can process a document with a single or multiple pages. To compare with similar applications, our application is processed very fast, without limitation of pages for each document and obtained accuracy as our expectation.","PeriodicalId":296963,"journal":{"name":"2021 8th NAFOSTED Conference on Information and Computer Science (NICS)","volume":"27 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2021-12-21","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"124449792","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2021-12-21DOI: 10.1109/NICS54270.2021.9701538
Minh-Trieu Tran, Bao V. Tran, Nguyen D. Vo, Khang Nguyen
In recent years, although various research has been performed on object detection with clear weather images, little attention has been paid to object detection with foggy aerial images. In this paper, we address the problem of detecting objects in foggy aerial images. Firstly, we create the UIT-DroneFog dataset by implementing a fog simulator (taken from the imgaug library) on 15,370 aerial images collected from the UIT-Drone21 dataset. This dataset has its distinguishing characteristic of having dense motorbike density in Vietnam with 4 objects: Pedestrian, Motor, Car, and Bus. Secondly, we further leverage two state-of-the-art object methods: Guided Anchoring, and Double Heads. The experiment results show that Double Heads achieve a higher mAP score, with 33.20%. Additionally, we propose a method called CasDou, which is the combination of Cascade RCNN, Double Heads, and Focal Loss. CasDou remarkably improves the mAP score up to 34.70%. The comprehensive evaluation points out the advantages and limitations of each method, which is the fundamental basement for further work.
{"title":"UIT-DroneFog: Toward High-performance Object Detection Via High-quality Aerial Foggy Dataset","authors":"Minh-Trieu Tran, Bao V. Tran, Nguyen D. Vo, Khang Nguyen","doi":"10.1109/NICS54270.2021.9701538","DOIUrl":"https://doi.org/10.1109/NICS54270.2021.9701538","url":null,"abstract":"In recent years, although various research has been performed on object detection with clear weather images, little attention has been paid to object detection with foggy aerial images. In this paper, we address the problem of detecting objects in foggy aerial images. Firstly, we create the UIT-DroneFog dataset by implementing a fog simulator (taken from the imgaug library) on 15,370 aerial images collected from the UIT-Drone21 dataset. This dataset has its distinguishing characteristic of having dense motorbike density in Vietnam with 4 objects: Pedestrian, Motor, Car, and Bus. Secondly, we further leverage two state-of-the-art object methods: Guided Anchoring, and Double Heads. The experiment results show that Double Heads achieve a higher mAP score, with 33.20%. Additionally, we propose a method called CasDou, which is the combination of Cascade RCNN, Double Heads, and Focal Loss. CasDou remarkably improves the mAP score up to 34.70%. The comprehensive evaluation points out the advantages and limitations of each method, which is the fundamental basement for further work.","PeriodicalId":296963,"journal":{"name":"2021 8th NAFOSTED Conference on Information and Computer Science (NICS)","volume":"38 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2021-12-21","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"116681128","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2021-12-21DOI: 10.1109/NICS54270.2021.9701454
Hoang-Viet Tran, Pham Ngoc Hung
Regression test is a well-known method to ensure that both unchanged and evolved functions of the evolving software are in good quality. This is known to be an expensive task even with automated test data generation methods. For this reason, this paper proposes an effective method to maintain and reuse the test data generation results from previous versions for regression test of the evolved version. The key idea of the proposed method is that for unchanged units, we can reuse the whole previous test data. For evolved units, we reuse as many as possible the solutions of the unchanged test path constraints and generate new test data for only the new or updated test path constraints. The analysis shows that the proposed method has a high potential of applicability in regression test of C/C++ projects in practice. We give discussions about several evolving scenarios of a given unit and how the proposed method comes to effective in such scenarios.
{"title":"An Efficient Method for Automated Regression Test Data Generation for C/C++ Projects","authors":"Hoang-Viet Tran, Pham Ngoc Hung","doi":"10.1109/NICS54270.2021.9701454","DOIUrl":"https://doi.org/10.1109/NICS54270.2021.9701454","url":null,"abstract":"Regression test is a well-known method to ensure that both unchanged and evolved functions of the evolving software are in good quality. This is known to be an expensive task even with automated test data generation methods. For this reason, this paper proposes an effective method to maintain and reuse the test data generation results from previous versions for regression test of the evolved version. The key idea of the proposed method is that for unchanged units, we can reuse the whole previous test data. For evolved units, we reuse as many as possible the solutions of the unchanged test path constraints and generate new test data for only the new or updated test path constraints. The analysis shows that the proposed method has a high potential of applicability in regression test of C/C++ projects in practice. We give discussions about several evolving scenarios of a given unit and how the proposed method comes to effective in such scenarios.","PeriodicalId":296963,"journal":{"name":"2021 8th NAFOSTED Conference on Information and Computer Science (NICS)","volume":"104 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2021-12-21","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"126162461","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2021-12-21DOI: 10.1109/NICS54270.2021.9701509
T. Trong, Quan-Tran Hai, Manh Vu Van, B. N. Thai, Tung Nguyen Chi, Truong Nguyen Quang
Real-time detection and tracking moving ships as well as locating the helipad on that ship is still a challenge in critical missions at sea. By using VTOL (Vertical Take-Off and Landing) UAV (Unmanned Aerial Vehicle) types, it allows both the ability to fly approaching the ship and also be able to land vertically on the helipad. This paper proposes a SITL (Software In The Loop) system to verify the automatic detection and tracking of ships moving at sea for the VTOL UAV during the mission and an algorithm to guide VTOL UAV mode selection in the process of approaching and landing on ships moving at sea. On-air images collected from the VTOL UAV’s camera in the X-Plane 11 simulation environment are used to train Deep Learning computer vision algorithms. Real-time ship detection algorithm with up to 125 FPS and 96% accuracy. From the results of the ship and helipad detection, we propose an algorithm to assist the transition of flight modes of VTOL UAV during the tracking and landing mission on a moving ship at sea.
在海上的关键任务中，实时探测和跟踪移动的船只以及定位船上的直升机停机坪仍然是一个挑战。通过使用VTOL(垂直起飞和降落)UAV(无人驾驶飞行器)类型，它允许两种能力飞接近舰船并且也能够垂直降落在直升机停机坪上。本文提出了一种用于验证垂直起降无人机在执行任务过程中对海上运动船舶自动检测和跟踪的SITL (Software In The Loop)系统，以及指导垂直起降无人机在接近和降落海上运动船舶过程中模式选择的算法。在X-Plane 11模拟环境中，垂直起降无人机的相机收集的空中图像用于训练深度学习计算机视觉算法。实时船舶检测算法，高达125 FPS和96%的精度。根据舰船和直升机停机坪的检测结果，提出了一种辅助垂直起降无人机在海上移动舰船跟踪降落任务中飞行模式转换的算法。
{"title":"Autonomous Detection and Approach Tracking of Moving Ship on the Sea by VTOL UAV based on Deep Learning Technique through Simulated Real-time On-Air Image Acquisitions","authors":"T. Trong, Quan-Tran Hai, Manh Vu Van, B. N. Thai, Tung Nguyen Chi, Truong Nguyen Quang","doi":"10.1109/NICS54270.2021.9701509","DOIUrl":"https://doi.org/10.1109/NICS54270.2021.9701509","url":null,"abstract":"Real-time detection and tracking moving ships as well as locating the helipad on that ship is still a challenge in critical missions at sea. By using VTOL (Vertical Take-Off and Landing) UAV (Unmanned Aerial Vehicle) types, it allows both the ability to fly approaching the ship and also be able to land vertically on the helipad. This paper proposes a SITL (Software In The Loop) system to verify the automatic detection and tracking of ships moving at sea for the VTOL UAV during the mission and an algorithm to guide VTOL UAV mode selection in the process of approaching and landing on ships moving at sea. On-air images collected from the VTOL UAV’s camera in the X-Plane 11 simulation environment are used to train Deep Learning computer vision algorithms. Real-time ship detection algorithm with up to 125 FPS and 96% accuracy. From the results of the ship and helipad detection, we propose an algorithm to assist the transition of flight modes of VTOL UAV during the tracking and landing mission on a moving ship at sea.","PeriodicalId":296963,"journal":{"name":"2021 8th NAFOSTED Conference on Information and Computer Science (NICS)","volume":"7 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2021-12-21","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"126018483","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2021-12-21DOI: 10.1109/NICS54270.2021.9701558
Duy A. Pham, A. D. Le, Dong T. Pham, H. B. Vo
Fruit flies become one of the most worrisome insect species to fruit yields. AlertTrap proposes and tests the constituent components to construct an efficient autonomous trap which sends notification to farmers when the number of flies exceeds a predefined threshold. The trap is powered with solar panels, equipped with a Lynfield-inspired sticky trap that is optimized to be attractive to fruit flies and controlled by an Arduino Board to collect data and circulate the energy through the system. The fruit flies are then counted on a Raspberry Pi Board by YOLOv4-tiny and SSD-MobileNet object detection algorithms with over 95% average precision at IoU threshold of 0.5 and an alert signal is sent to the farmers based on the number of fruit flies in the trap.
{"title":"AlertTrap: On Designing An Edge-Computing Remote Insect Monitoring System","authors":"Duy A. Pham, A. D. Le, Dong T. Pham, H. B. Vo","doi":"10.1109/NICS54270.2021.9701558","DOIUrl":"https://doi.org/10.1109/NICS54270.2021.9701558","url":null,"abstract":"Fruit flies become one of the most worrisome insect species to fruit yields. AlertTrap proposes and tests the constituent components to construct an efficient autonomous trap which sends notification to farmers when the number of flies exceeds a predefined threshold. The trap is powered with solar panels, equipped with a Lynfield-inspired sticky trap that is optimized to be attractive to fruit flies and controlled by an Arduino Board to collect data and circulate the energy through the system. The fruit flies are then counted on a Raspberry Pi Board by YOLOv4-tiny and SSD-MobileNet object detection algorithms with over 95% average precision at IoU threshold of 0.5 and an alert signal is sent to the farmers based on the number of fruit flies in the trap.","PeriodicalId":296963,"journal":{"name":"2021 8th NAFOSTED Conference on Information and Computer Science (NICS)","volume":"81 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2021-12-21","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"133602067","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
This paper proposes a highly digital Analog-to-Digital Converter (ADC) for various internet of things applications, which operates at the bandwidth below 50 kHz. Our goal is a highly digital ADC that can be integrated into the low-power System-on-Chip (SoC) to adapt to IoT demands like audio recording or sensor measuring. The ADC is implemented using only ring-oscillator and digital circuits that is based on time-encoding technique and Delta-Sigma modulation. In this work, we optimize the Voltage Control Oscillator (VCO) for high linearity and apply a Cascaded Integrator Comb (CIC) filter with the aim of increasing the ADC’s resolution. Our work is implemented and verified by fully open-source tools on the Skywater 130 - nm technology. The ADC produces more significant than 12 effective bits at the cost of 0.97 mW and occupies 0.08 mm2.
{"title":"A Highly Digital VCO-based ADC for IoT Applications on Skywater 130nm","authors":"Duc-Manh Tran, Ngo-Doanh Nguyen, Duy-Hieu Bui, Xuan-Tu Tran","doi":"10.1109/NICS54270.2021.9701515","DOIUrl":"https://doi.org/10.1109/NICS54270.2021.9701515","url":null,"abstract":"This paper proposes a highly digital Analog-to-Digital Converter (ADC) for various internet of things applications, which operates at the bandwidth below 50 kHz. Our goal is a highly digital ADC that can be integrated into the low-power System-on-Chip (SoC) to adapt to IoT demands like audio recording or sensor measuring. The ADC is implemented using only ring-oscillator and digital circuits that is based on time-encoding technique and Delta-Sigma modulation. In this work, we optimize the Voltage Control Oscillator (VCO) for high linearity and apply a Cascaded Integrator Comb (CIC) filter with the aim of increasing the ADC’s resolution. Our work is implemented and verified by fully open-source tools on the Skywater 130 - nm technology. The ADC produces more significant than 12 effective bits at the cost of 0.97 mW and occupies 0.08 mm2.","PeriodicalId":296963,"journal":{"name":"2021 8th NAFOSTED Conference on Information and Computer Science (NICS)","volume":"1 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2021-12-21","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"131277830","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2021-12-21DOI: 10.1109/NICS54270.2021.9701549
Hieu Trung Nguyen, Khang Tran, N. H. Luong
Hybridizations of Deep Reinforcement Learning (DRL) and Evolution Computation (EC) methods have recently showed considerable successes in a variety of high dimensional physical control tasks. These hybrid frameworks offer more robust mechanisms of exploration and exploitation in the policy network parameter search space when stabilizing gradient-based updates of DRL algorithms with population-based operations adopted from EC methods. In this paper, we propose a novel hybrid framework that effectively combines the efficiency of DRL updates and the stability of EC populations. We experiment with integrating the Twin Delayed Deep Deterministic Policy Gradient (TD3) and the Cross-Entropy Method (CEM). The resulting EC-enhanced TD3 algorithm (eTD3) are compared with the baseline algorithm TD3 and a state-of-the-art evolutionary reinforcement learning (ERL) method, CEM-TD3. Experimental results on five MuJoCo continuous control benchmark environments confirm the efficacy of our approach. The source code of the paper is available at https://github.com/ELO-Lab/eTD3.
{"title":"Enhancing Twin Delayed Deep Deterministic Policy Gradient with Cross-Entropy Method","authors":"Hieu Trung Nguyen, Khang Tran, N. H. Luong","doi":"10.1109/NICS54270.2021.9701549","DOIUrl":"https://doi.org/10.1109/NICS54270.2021.9701549","url":null,"abstract":"Hybridizations of Deep Reinforcement Learning (DRL) and Evolution Computation (EC) methods have recently showed considerable successes in a variety of high dimensional physical control tasks. These hybrid frameworks offer more robust mechanisms of exploration and exploitation in the policy network parameter search space when stabilizing gradient-based updates of DRL algorithms with population-based operations adopted from EC methods. In this paper, we propose a novel hybrid framework that effectively combines the efficiency of DRL updates and the stability of EC populations. We experiment with integrating the Twin Delayed Deep Deterministic Policy Gradient (TD3) and the Cross-Entropy Method (CEM). The resulting EC-enhanced TD3 algorithm (eTD3) are compared with the baseline algorithm TD3 and a state-of-the-art evolutionary reinforcement learning (ERL) method, CEM-TD3. Experimental results on five MuJoCo continuous control benchmark environments confirm the efficacy of our approach. The source code of the paper is available at https://github.com/ELO-Lab/eTD3.","PeriodicalId":296963,"journal":{"name":"2021 8th NAFOSTED Conference on Information and Computer Science (NICS)","volume":"1 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2021-12-21","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"131327430","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2021-12-21DOI: 10.1109/NICS54270.2021.9701461
K. Nurhanim, I. Elamvazuthi, L. I. Izhar, G. Capi, Steven W. Su
In Human activity recognition (HAR) research, it is a common practice to use wearable sensors to acquire the signals for human daily activities. In this study, an experimental data from electromyography (EMG) wireless sensors is analysed for six different activities recognition. This paper aims to compare EMG signals of left and right of upper leg muscles by using Random Forest (RF) Machine Learning Classifier. The HAR processing comprises of data filtering and segmentation, data feature extraction, feature selection of the data, and classification. Model evaluation of holdout method is implemented for classification assessment. The performance of all human daily activities is evaluated according to the comparison of precision and recall for each activity. The results show combined muscles obtained the highest precision and recall on running activity with 89.2% and 88.3%. The highest overall accuracy of classification was 82.08% on the bicep femoris left and right (BF-Left & Right).
{"title":"EMG Signals Classification on Human Activity Recognition using Machine Learning Algorithm","authors":"K. Nurhanim, I. Elamvazuthi, L. I. Izhar, G. Capi, Steven W. Su","doi":"10.1109/NICS54270.2021.9701461","DOIUrl":"https://doi.org/10.1109/NICS54270.2021.9701461","url":null,"abstract":"In Human activity recognition (HAR) research, it is a common practice to use wearable sensors to acquire the signals for human daily activities. In this study, an experimental data from electromyography (EMG) wireless sensors is analysed for six different activities recognition. This paper aims to compare EMG signals of left and right of upper leg muscles by using Random Forest (RF) Machine Learning Classifier. The HAR processing comprises of data filtering and segmentation, data feature extraction, feature selection of the data, and classification. Model evaluation of holdout method is implemented for classification assessment. The performance of all human daily activities is evaluated according to the comparison of precision and recall for each activity. The results show combined muscles obtained the highest precision and recall on running activity with 89.2% and 88.3%. The highest overall accuracy of classification was 82.08% on the bicep femoris left and right (BF-Left & Right).","PeriodicalId":296963,"journal":{"name":"2021 8th NAFOSTED Conference on Information and Computer Science (NICS)","volume":"33 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2021-12-21","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"134058413","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
With the rapid development of information and technology, document digitization has become more critical in many research fields by giving enormous amounts of data. However, computers can not handle a lot of information contained inside physical documents. For that reason, making computers detect objects in document images can help humans have more valuable information such as graphs, captions, or tables. There should be a system capable of detecting various components on document images, especially finding a simply effective object recognition method. Thus, the introduction of YOLOF can be an appropriate method to detect objects in documents because it opens up a simple way to exploit image features, making the object detection problem less computationally intensive, but still maintaining the appropriate accuracy. This paper evaluates the new one-stage YOLOF method on two challenging document datasets: IIIT-AR-13K, UIT-DODV. Our experimental YOLOF model achieves 58.8% and 56% on mAP measurement scores with the IIIT-AR-13K dataset and the UIT-DODV dataset, respectively.
{"title":"Page Object Detection with YOLOF","authors":"Phuc Nguyen, Luu Ngo, Thang Truong, Trong-Thuan Nguyen, Nguyen D. Vo, Khang Nguyen","doi":"10.1109/NICS54270.2021.9701449","DOIUrl":"https://doi.org/10.1109/NICS54270.2021.9701449","url":null,"abstract":"With the rapid development of information and technology, document digitization has become more critical in many research fields by giving enormous amounts of data. However, computers can not handle a lot of information contained inside physical documents. For that reason, making computers detect objects in document images can help humans have more valuable information such as graphs, captions, or tables. There should be a system capable of detecting various components on document images, especially finding a simply effective object recognition method. Thus, the introduction of YOLOF can be an appropriate method to detect objects in documents because it opens up a simple way to exploit image features, making the object detection problem less computationally intensive, but still maintaining the appropriate accuracy. This paper evaluates the new one-stage YOLOF method on two challenging document datasets: IIIT-AR-13K, UIT-DODV. Our experimental YOLOF model achieves 58.8% and 56% on mAP measurement scores with the IIIT-AR-13K dataset and the UIT-DODV dataset, respectively.","PeriodicalId":296963,"journal":{"name":"2021 8th NAFOSTED Conference on Information and Computer Science (NICS)","volume":"56 80 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2021-12-21","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"129341755","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2021-12-21DOI: 10.1109/nics54270.2021.9701579
{"title":"[Copyright notice]","authors":"","doi":"10.1109/nics54270.2021.9701579","DOIUrl":"https://doi.org/10.1109/nics54270.2021.9701579","url":null,"abstract":"","PeriodicalId":296963,"journal":{"name":"2021 8th NAFOSTED Conference on Information and Computer Science (NICS)","volume":"82 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2021-12-21","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"116152205","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
京公网安备 11010802042870号
京ICP备2023020795号-1
扫码关注我们
求助内容：
应助结果提醒方式：