Pub Date : 2020-07-01DOI: 10.1109/ICIVC50857.2020.9177480
Jiating Jin, Xingwei Li, Xinlong Li, Shaojie Guan
Simple Online and Realtime Tracking with a Deep Association Metric (DeepSORT) is a method of multi-object tracking combined appearance features with motion state of objects estimated by Kalman Filter which has a promising performance. However, maintaining the identity of targets becomes formidable when the objects have a similar appearance and complex patterns of the movement. To address these issues, a novel Online Multi-object Tracking with Siamese Network and Optical Flow is proposed. We utilize the Siamese network structure to obtain our appearance feature extractor. Furthermore, optical flow is introduced into the scheme to promote the accuracy of motion prediction from the Kalman filter. Our approach combines appearance and motion features in a tracking framework. The experimental results evaluated on the public MOT dataset illustrate that our method has the better performance in comparison with the DeepSORT algorithm.
{"title":"Online Multi-object Tracking with Siamese Network and Optical Flow","authors":"Jiating Jin, Xingwei Li, Xinlong Li, Shaojie Guan","doi":"10.1109/ICIVC50857.2020.9177480","DOIUrl":"https://doi.org/10.1109/ICIVC50857.2020.9177480","url":null,"abstract":"Simple Online and Realtime Tracking with a Deep Association Metric (DeepSORT) is a method of multi-object tracking combined appearance features with motion state of objects estimated by Kalman Filter which has a promising performance. However, maintaining the identity of targets becomes formidable when the objects have a similar appearance and complex patterns of the movement. To address these issues, a novel Online Multi-object Tracking with Siamese Network and Optical Flow is proposed. We utilize the Siamese network structure to obtain our appearance feature extractor. Furthermore, optical flow is introduced into the scheme to promote the accuracy of motion prediction from the Kalman filter. Our approach combines appearance and motion features in a tracking framework. The experimental results evaluated on the public MOT dataset illustrate that our method has the better performance in comparison with the DeepSORT algorithm.","PeriodicalId":6806,"journal":{"name":"2020 IEEE 5th International Conference on Image, Vision and Computing (ICIVC)","volume":"5 1","pages":"193-198"},"PeriodicalIF":0.0,"publicationDate":"2020-07-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"72830449","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 : 2020-07-01DOI: 10.1109/ICIVC50857.2020.9177438
Jinda Hu, Yanshun Zhao, Xindong Zhang
Object detection is one of the most important and challenging branches of computer vision, which has been widely applied in our life. With the rapid development of deep learning networks for detection tasks, the performance of object detectors has been greatly improved. However, the lack of large labeled dataset obstructs the usage of convolutional neural networks (CNN) for detecting in thermal infrared images. Most existing dataset focus on visible images, while thermal infrared images are helpful for detection even in a dark environment. To address this problem, we propose the use of transfer learning to improve the accuracy of infrared pedestrian detection. We pretrain a convolutional neural network on a large dataset (which contains 1.8 million images with 654 categories), then use the convolutional neural network as a fixed feature extractor for the task of infrared pedestrian detection. The average precision of detection using ImageNet pretrained model alone is 83.34%. By adding ours pretrained model, the average precision has improved to 84.78%. We believe that the method of transfer learning can be extended to other infrared detection applications and achieve other breakthroughs.
{"title":"Application of Transfer Learning in Infrared Pedestrian Detection","authors":"Jinda Hu, Yanshun Zhao, Xindong Zhang","doi":"10.1109/ICIVC50857.2020.9177438","DOIUrl":"https://doi.org/10.1109/ICIVC50857.2020.9177438","url":null,"abstract":"Object detection is one of the most important and challenging branches of computer vision, which has been widely applied in our life. With the rapid development of deep learning networks for detection tasks, the performance of object detectors has been greatly improved. However, the lack of large labeled dataset obstructs the usage of convolutional neural networks (CNN) for detecting in thermal infrared images. Most existing dataset focus on visible images, while thermal infrared images are helpful for detection even in a dark environment. To address this problem, we propose the use of transfer learning to improve the accuracy of infrared pedestrian detection. We pretrain a convolutional neural network on a large dataset (which contains 1.8 million images with 654 categories), then use the convolutional neural network as a fixed feature extractor for the task of infrared pedestrian detection. The average precision of detection using ImageNet pretrained model alone is 83.34%. By adding ours pretrained model, the average precision has improved to 84.78%. We believe that the method of transfer learning can be extended to other infrared detection applications and achieve other breakthroughs.","PeriodicalId":6806,"journal":{"name":"2020 IEEE 5th International Conference on Image, Vision and Computing (ICIVC)","volume":"26 1","pages":"1-4"},"PeriodicalIF":0.0,"publicationDate":"2020-07-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"79358495","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 : 2020-07-01DOI: 10.1109/ICIVC50857.2020.9177446
Peitong Li, Qiuju Deng, Huiqi Li
Retinal vessel is the only vessel structure in human circulatory system that can be directly observed by non-invasive methods. According to clinical findings, the reduction of arteriovenous width ratio (AVR) acts as an indicator to predict the risk of many systemic diseases. Therefore, it's essential to develop an automatic classification method for arteries and veins to calculate AVR. A method that combines the deep segmentation network and tracking algorithm is proposed in this paper to classify arteries and veins in retinal images. This automatic processing has three steps: (1) retinal images are preprocessed with a haze-removal technique (2) a U-net segmentation network is utilized to classify pixels into background, artery or vein (3) a tracking algorithm is applied for vessel-wise classifications. The proposed method is tested on a clinical dataset and the results present an accuracy of 93.57% for vessel-wise classifications.
{"title":"The Arteriovenous Classification in Retinal Images by U-net and Tracking Algorithm","authors":"Peitong Li, Qiuju Deng, Huiqi Li","doi":"10.1109/ICIVC50857.2020.9177446","DOIUrl":"https://doi.org/10.1109/ICIVC50857.2020.9177446","url":null,"abstract":"Retinal vessel is the only vessel structure in human circulatory system that can be directly observed by non-invasive methods. According to clinical findings, the reduction of arteriovenous width ratio (AVR) acts as an indicator to predict the risk of many systemic diseases. Therefore, it's essential to develop an automatic classification method for arteries and veins to calculate AVR. A method that combines the deep segmentation network and tracking algorithm is proposed in this paper to classify arteries and veins in retinal images. This automatic processing has three steps: (1) retinal images are preprocessed with a haze-removal technique (2) a U-net segmentation network is utilized to classify pixels into background, artery or vein (3) a tracking algorithm is applied for vessel-wise classifications. The proposed method is tested on a clinical dataset and the results present an accuracy of 93.57% for vessel-wise classifications.","PeriodicalId":6806,"journal":{"name":"2020 IEEE 5th International Conference on Image, Vision and Computing (ICIVC)","volume":"45 1","pages":"182-187"},"PeriodicalIF":0.0,"publicationDate":"2020-07-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"75583024","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 : 2020-07-01DOI: 10.1109/ICIVC50857.2020.9177483
Erqian Cai
Image is one of the core concerns in modern Astronomy. Telescopes capture photons emitted from sources deep inside the universe, forming images or spectrums which then be analyzed by astronomers. In the recent decades, people have built large amount of land-based and space-based telescopes which are observing light covering a wide range of wave length. The amount of the imaging data increased rapidly. For a typical integral field unit (also called the IFU) telescope, 60 GB of data is generated each night. The requirements of real time processing of these data raised challenges to astronomers. These requirements necessitate the developing of efficient computer algorithms. One important part of these requirements is the classification of galaxies. The morphologies of the galaxies can contribute in many aspects of the astronomical studies. The distribution of galaxies of different morphologies (for example ecliptic and spiral) can reflect certain large scale characteristic of the universe, such as the evolution of the galaxies, and the distribution of Hydrogen in the universe. In this work, we train a neural network and use a series of computer vision algorithms to build a Galaxy Detection and Classification Tool (GalaDC), which can detect and classify galaxies with high efficiency and accuracy. GalaDC is user friendly, supports batch processing, and is suitable for handling images which consists of multiple galaxies and do statistical analysis.
{"title":"GalaDC: Galaxy Detection and Classification Tool","authors":"Erqian Cai","doi":"10.1109/ICIVC50857.2020.9177483","DOIUrl":"https://doi.org/10.1109/ICIVC50857.2020.9177483","url":null,"abstract":"Image is one of the core concerns in modern Astronomy. Telescopes capture photons emitted from sources deep inside the universe, forming images or spectrums which then be analyzed by astronomers. In the recent decades, people have built large amount of land-based and space-based telescopes which are observing light covering a wide range of wave length. The amount of the imaging data increased rapidly. For a typical integral field unit (also called the IFU) telescope, 60 GB of data is generated each night. The requirements of real time processing of these data raised challenges to astronomers. These requirements necessitate the developing of efficient computer algorithms. One important part of these requirements is the classification of galaxies. The morphologies of the galaxies can contribute in many aspects of the astronomical studies. The distribution of galaxies of different morphologies (for example ecliptic and spiral) can reflect certain large scale characteristic of the universe, such as the evolution of the galaxies, and the distribution of Hydrogen in the universe. In this work, we train a neural network and use a series of computer vision algorithms to build a Galaxy Detection and Classification Tool (GalaDC), which can detect and classify galaxies with high efficiency and accuracy. GalaDC is user friendly, supports batch processing, and is suitable for handling images which consists of multiple galaxies and do statistical analysis.","PeriodicalId":6806,"journal":{"name":"2020 IEEE 5th International Conference on Image, Vision and Computing (ICIVC)","volume":"30 1","pages":"261-266"},"PeriodicalIF":0.0,"publicationDate":"2020-07-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"78161441","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 : 2020-07-01DOI: 10.1109/ICIVC50857.2020.9177451
Chao Wang, Zeyu Gao, Qiankun Lu
As an effective image preprocessing method, color correction is widely used in the field of image stitching. However, due to the changeable visual environment, parallax often occurs in mosaic images. So far, there are few algorithms dealing with color correction under parallax situation. And most color correction methods produce poor results in images with parallax. This paper introduces a color correction algorithm which can also be applied in the case of parallax in image mosaic. We use VFC algorithm, color hue information and iterative calculation to obtain better color correction results. Compared with existing algorithms, it has the wider range of applications. The experiments show that our method has faster time efficiency and better results.
{"title":"Parallax-Based Color Correction in Image Stitching","authors":"Chao Wang, Zeyu Gao, Qiankun Lu","doi":"10.1109/ICIVC50857.2020.9177451","DOIUrl":"https://doi.org/10.1109/ICIVC50857.2020.9177451","url":null,"abstract":"As an effective image preprocessing method, color correction is widely used in the field of image stitching. However, due to the changeable visual environment, parallax often occurs in mosaic images. So far, there are few algorithms dealing with color correction under parallax situation. And most color correction methods produce poor results in images with parallax. This paper introduces a color correction algorithm which can also be applied in the case of parallax in image mosaic. We use VFC algorithm, color hue information and iterative calculation to obtain better color correction results. Compared with existing algorithms, it has the wider range of applications. The experiments show that our method has faster time efficiency and better results.","PeriodicalId":6806,"journal":{"name":"2020 IEEE 5th International Conference on Image, Vision and Computing (ICIVC)","volume":"33 1","pages":"69-74"},"PeriodicalIF":0.0,"publicationDate":"2020-07-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"76182410","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 : 2020-07-01DOI: 10.1109/ICIVC50857.2020.9177466
Jie Zhou, Rui Yang, Mengying Xu
In order to effectively improve the efficiency of multi-constrained QoS routing and reduce the energy consumption of data on the transmission path an efficient routing algorithm needs to be designed. Aiming at the problem of constrained QoS routing, an adaptive chaotic shuffled frog leaping algorithm is designed, a graph theory model of wireless image sensor network is established, and a corresponding fitness function is derived to find the path with the least energy consumption. Added new adaptive operator and chaotic operator to improve the global search ability. In the simulation, the adaptive chaotic shuffled frog leap algorithm is compared with evolutionary algorithm and particle swarm optimization. The experimental results prove that compared with evolutionary algorithm and particle swarm optimization the adaptive chaotic shuffled frog leap algorithm can be effectively accelerate convergence speed and reduce the energy loss of data on the transmission path.
{"title":"Adaptive Chaotic Shuffled Frog Leaping Algorithm for QoS Routing in Wireless Image Sensor Networks","authors":"Jie Zhou, Rui Yang, Mengying Xu","doi":"10.1109/ICIVC50857.2020.9177466","DOIUrl":"https://doi.org/10.1109/ICIVC50857.2020.9177466","url":null,"abstract":"In order to effectively improve the efficiency of multi-constrained QoS routing and reduce the energy consumption of data on the transmission path an efficient routing algorithm needs to be designed. Aiming at the problem of constrained QoS routing, an adaptive chaotic shuffled frog leaping algorithm is designed, a graph theory model of wireless image sensor network is established, and a corresponding fitness function is derived to find the path with the least energy consumption. Added new adaptive operator and chaotic operator to improve the global search ability. In the simulation, the adaptive chaotic shuffled frog leap algorithm is compared with evolutionary algorithm and particle swarm optimization. The experimental results prove that compared with evolutionary algorithm and particle swarm optimization the adaptive chaotic shuffled frog leap algorithm can be effectively accelerate convergence speed and reduce the energy loss of data on the transmission path.","PeriodicalId":6806,"journal":{"name":"2020 IEEE 5th International Conference on Image, Vision and Computing (ICIVC)","volume":"21 1","pages":"287-291"},"PeriodicalIF":0.0,"publicationDate":"2020-07-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"81797110","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 : 2020-07-01DOI: 10.1109/ICIVC50857.2020.9177432
Chen Fangfang, Cao Peng
Ordinary QR code does not have the function of anti duplication, and it is easy to be copied and counterfeited. In order to overcome this loophole, an anti-counterfeiting method using CMYK color printing metamerism characteristics + embedded pseudo-random noise is adopted to generate a QR code with dual anti duplication and anti-counterfeiting (secure QR code for short). Firstly, the anti-counterfeiting information is encoded and embedded into the ordinary QR code in the form of noise to generate the first-level anti-copying and anti-counterfeiting QR code with information. Then the CMYK color ratio is modulated with the second-level anti-counterfeiting information to satisfy the metamerism characteristic, and the second-level anti-counterfeiting is realized. Experimental tests show that the secure QR code designed in this paper has good information hiding performance, anti-replication performance and multi-code integration function, and can be used in anti-counterfeiting of trademarks, books, labels, etc.
{"title":"Research on Dual Anti Duplication and Anti-counterfeiting Technology of QR Code Based on Metamerism Characteristics","authors":"Chen Fangfang, Cao Peng","doi":"10.1109/ICIVC50857.2020.9177432","DOIUrl":"https://doi.org/10.1109/ICIVC50857.2020.9177432","url":null,"abstract":"Ordinary QR code does not have the function of anti duplication, and it is easy to be copied and counterfeited. In order to overcome this loophole, an anti-counterfeiting method using CMYK color printing metamerism characteristics + embedded pseudo-random noise is adopted to generate a QR code with dual anti duplication and anti-counterfeiting (secure QR code for short). Firstly, the anti-counterfeiting information is encoded and embedded into the ordinary QR code in the form of noise to generate the first-level anti-copying and anti-counterfeiting QR code with information. Then the CMYK color ratio is modulated with the second-level anti-counterfeiting information to satisfy the metamerism characteristic, and the second-level anti-counterfeiting is realized. Experimental tests show that the secure QR code designed in this paper has good information hiding performance, anti-replication performance and multi-code integration function, and can be used in anti-counterfeiting of trademarks, books, labels, etc.","PeriodicalId":6806,"journal":{"name":"2020 IEEE 5th International Conference on Image, Vision and Computing (ICIVC)","volume":"14 1","pages":"303-306"},"PeriodicalIF":0.0,"publicationDate":"2020-07-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"87142358","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 : 2020-07-01DOI: 10.1109/ICIVC50857.2020.9177458
Weijun Zhai
The ever-changing external economic environment brings new challenges to the audit supervision department to strengthen the management of audit failure. This paper analyzes the relationship between audit failure and the mechanism of stimulation and punishment based on the game theory analysis model of managers' behavior in project management. In order to avoid the audit failure, this paper provides effective analysis tools and suggestions to improve the mechanism of current incentive and punishment in the final.
{"title":"Research on the Audit Failure Based on the Perspective of Manager Behavior Game","authors":"Weijun Zhai","doi":"10.1109/ICIVC50857.2020.9177458","DOIUrl":"https://doi.org/10.1109/ICIVC50857.2020.9177458","url":null,"abstract":"The ever-changing external economic environment brings new challenges to the audit supervision department to strengthen the management of audit failure. This paper analyzes the relationship between audit failure and the mechanism of stimulation and punishment based on the game theory analysis model of managers' behavior in project management. In order to avoid the audit failure, this paper provides effective analysis tools and suggestions to improve the mechanism of current incentive and punishment in the final.","PeriodicalId":6806,"journal":{"name":"2020 IEEE 5th International Conference on Image, Vision and Computing (ICIVC)","volume":"28 1","pages":"307-311"},"PeriodicalIF":0.0,"publicationDate":"2020-07-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"87410375","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 : 2020-07-01DOI: 10.1109/ICIVC50857.2020.9177443
M. Guarneri, M. F. de Collibus, M. Francucci, M. Ciaffi
At the moment when this article is written, a pandemic disease is attacking our lives, our style of living and our economy. The present work uses this occasion for focusing the attention on the importance to make available a digital copy of our knowledge, history and habits. The slower passing of time inside own residence let the individual to rediscover natural indoor activities, like reading a book or watching a documentary, and try to mentally escape by a virtual visit in a museum or a city. The first evidence coming out from these sites is mainly the limits of this technology for appreciating the artworks, even inside 3D environments, and, probably the most important, the lack of standardization in terms of accessibility and quality of the products. The present work focuses the attention only on one of the aspects of the processes for studying and documenting an artwork: the data acquisition and preprocessing data fusion. For approaching these steps, an out-of-the-market 3D technology based on the combination of several laser sources will be described: the description of this kind of systems is the pretext for analyzing the main differences with the available devices and techniques today largely used in Cultural Heritage environment, but especially for highlighting how the research can try to unify the gamification with diagnostic and restoration support in this sector.
{"title":"The Importance of Artworks 3D Digitalization at the Time of COVID Epidemy: Case Studies by the Use of a Multi-wavelengths Technique","authors":"M. Guarneri, M. F. de Collibus, M. Francucci, M. Ciaffi","doi":"10.1109/ICIVC50857.2020.9177443","DOIUrl":"https://doi.org/10.1109/ICIVC50857.2020.9177443","url":null,"abstract":"At the moment when this article is written, a pandemic disease is attacking our lives, our style of living and our economy. The present work uses this occasion for focusing the attention on the importance to make available a digital copy of our knowledge, history and habits. The slower passing of time inside own residence let the individual to rediscover natural indoor activities, like reading a book or watching a documentary, and try to mentally escape by a virtual visit in a museum or a city. The first evidence coming out from these sites is mainly the limits of this technology for appreciating the artworks, even inside 3D environments, and, probably the most important, the lack of standardization in terms of accessibility and quality of the products. The present work focuses the attention only on one of the aspects of the processes for studying and documenting an artwork: the data acquisition and preprocessing data fusion. For approaching these steps, an out-of-the-market 3D technology based on the combination of several laser sources will be described: the description of this kind of systems is the pretext for analyzing the main differences with the available devices and techniques today largely used in Cultural Heritage environment, but especially for highlighting how the research can try to unify the gamification with diagnostic and restoration support in this sector.","PeriodicalId":6806,"journal":{"name":"2020 IEEE 5th International Conference on Image, Vision and Computing (ICIVC)","volume":"219 1","pages":"113-117"},"PeriodicalIF":0.0,"publicationDate":"2020-07-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"88078535","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 : 2020-07-01DOI: 10.1109/ICIVC50857.2020.9177431
Jiawei Hou, Zhaohui Wang, Yigan Li
Makeup, derived from the human pursuit of beauty, it changes the image of people appearance, brings more beautiful enjoyment and spiritual pleasure. However, recent studies have shown that facial makeup have a negative effect on face verification. To solve this problem, we formulate an end-to-end deep learning network which is composed of a stem CNN and a novel mapping module. Specifically, we pre-train our framework on a comprehensive dataset and fine-tune our mapping module on makeup datasets. Then we experimentally validate the proposal on these datasets. Experimental results demonstrate that the proposal achieves promising performance compared to the existing state-of-the-art methods.
{"title":"A Network for Makeup Face Verification Based upon Deep Learning","authors":"Jiawei Hou, Zhaohui Wang, Yigan Li","doi":"10.1109/ICIVC50857.2020.9177431","DOIUrl":"https://doi.org/10.1109/ICIVC50857.2020.9177431","url":null,"abstract":"Makeup, derived from the human pursuit of beauty, it changes the image of people appearance, brings more beautiful enjoyment and spiritual pleasure. However, recent studies have shown that facial makeup have a negative effect on face verification. To solve this problem, we formulate an end-to-end deep learning network which is composed of a stem CNN and a novel mapping module. Specifically, we pre-train our framework on a comprehensive dataset and fine-tune our mapping module on makeup datasets. Then we experimentally validate the proposal on these datasets. Experimental results demonstrate that the proposal achieves promising performance compared to the existing state-of-the-art methods.","PeriodicalId":6806,"journal":{"name":"2020 IEEE 5th International Conference on Image, Vision and Computing (ICIVC)","volume":"57 1","pages":"123-127"},"PeriodicalIF":0.0,"publicationDate":"2020-07-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"90760232","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：