Pub Date : 2022-08-17DOI: 10.1109/IWIS56333.2022.9920903
Elena Filonenko, A. Filonenko, K. Jo
Recognition of text on low-resolution (LR) images is a challenging task. Traditional interpolation methods, as well as general super-resolution approaches, do not recover the shape of text character robustly. In this work, we propose a text-aware super- resolution neural network called TASuRe. Text awareness is interwoven in the proposed network that contains a text rectification part and text recognition auxiliary module. The training procedure is built around character shape restoration by adding a binary mask to the input image and using a specialized loss that penalties the network for missing gradients on the border of characters. Experiments on the real LR images have shown that the proposed network can deal with hard cases better than convolutional competitors.
{"title":"TASuRe: Text Aware Super-Resolution","authors":"Elena Filonenko, A. Filonenko, K. Jo","doi":"10.1109/IWIS56333.2022.9920903","DOIUrl":"https://doi.org/10.1109/IWIS56333.2022.9920903","url":null,"abstract":"Recognition of text on low-resolution (LR) images is a challenging task. Traditional interpolation methods, as well as general super-resolution approaches, do not recover the shape of text character robustly. In this work, we propose a text-aware super- resolution neural network called TASuRe. Text awareness is interwoven in the proposed network that contains a text rectification part and text recognition auxiliary module. The training procedure is built around character shape restoration by adding a binary mask to the input image and using a specialized loss that penalties the network for missing gradients on the border of characters. Experiments on the real LR images have shown that the proposed network can deal with hard cases better than convolutional competitors.","PeriodicalId":340399,"journal":{"name":"2022 International Workshop on Intelligent Systems (IWIS)","volume":"147 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2022-08-17","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"115277824","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 : 2022-08-17DOI: 10.1109/IWIS56333.2022.9920723
Hidenao Takahashi, Nami Takino, H. Hashimoto, D. Chugo, S. Muramatsu, S. Yokota, Jin-Hua She, H. Hashimoto
This paper proposes a new method for estimating the attributes of a pedestrian, e.g. an office worker or a student, from the pedestrian's clothing and, based on the estimation results, a method for efficiently navigating an autonomous mobile robot while safely avoiding pedestrians. The robot can estimate the organization the pedestrians belong to by learning in advance the clothing of pedestrians that frequently appear in the city in which the is travelling, e.g. the uniforms of certain high school students. Furthermore, the robot is given the direction in which the person in that organization is likely to be heading at the current time, so it uses this information to estimate the future walking trajectory of that pedestrian and design a running path that is less likely to prevent that pedestrian from walking. We use the artificial potential method to design the robot's running trajectory. Even if pedestrians with different attributes are detected in the robot's running path, our proposed robot can efficiently reach its destination while avoiding those pedestrians appropriately in real time. The effectiveness of the proposed method was confirmed by computer simulations and subjects' experiments with our prototype robot.
{"title":"Attribute estimation using pedestrian clothing and autonomous mobile robot navigation based on it","authors":"Hidenao Takahashi, Nami Takino, H. Hashimoto, D. Chugo, S. Muramatsu, S. Yokota, Jin-Hua She, H. Hashimoto","doi":"10.1109/IWIS56333.2022.9920723","DOIUrl":"https://doi.org/10.1109/IWIS56333.2022.9920723","url":null,"abstract":"This paper proposes a new method for estimating the attributes of a pedestrian, e.g. an office worker or a student, from the pedestrian's clothing and, based on the estimation results, a method for efficiently navigating an autonomous mobile robot while safely avoiding pedestrians. The robot can estimate the organization the pedestrians belong to by learning in advance the clothing of pedestrians that frequently appear in the city in which the is travelling, e.g. the uniforms of certain high school students. Furthermore, the robot is given the direction in which the person in that organization is likely to be heading at the current time, so it uses this information to estimate the future walking trajectory of that pedestrian and design a running path that is less likely to prevent that pedestrian from walking. We use the artificial potential method to design the robot's running trajectory. Even if pedestrians with different attributes are detected in the robot's running path, our proposed robot can efficiently reach its destination while avoiding those pedestrians appropriately in real time. The effectiveness of the proposed method was confirmed by computer simulations and subjects' experiments with our prototype robot.","PeriodicalId":340399,"journal":{"name":"2022 International Workshop on Intelligent Systems (IWIS)","volume":"48 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2022-08-17","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"123064458","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 : 2022-08-17DOI: 10.1109/IWIS56333.2022.9920917
Tan-Thien-Nien Nguyen, Thanh-Danh Phan, Minh-Thien Duong, Chi-Tam Nguyen, Hong-Phong Ly, M. Le
Obstacle dodging and overtaking are the pivotal tasks ensuring safety for self-driving automobiles. Multi-sensors fusion is the must-required condition to explore the entire surrounding information. This paper proposes a novel frontal dynamic car dodging strategy for automobiles with the left-hand side steering wheel by fusion of a camera and 2D LiDAR features. To begin with, we improve the LiteSeg model to extract the segmented mask, which can determine the drivable area and the avoiding direction. In addition to a camera, 2D LiDAR is used to check the scene information on the right side, which the camera's range cannot cover. As for point clouds output of 2D LiDAR, we adopt the Adaptive Breakpoint Detection (defined as ABD) algorithm to cluster the objects in a scanning plane. Subsequently, the RANSAC algorithm forms a straight line from the clustered point clouds to determine the boundary of the right-side obstacle. Besides, we compute the distance from LiDAR to the estimated straight line to maintain a safe distance when overtaking. Last but not least, the post-processing results of the two devices are fused to decide on the obstacle dodging and overtaking. The comprehensive experiments reveal that our self-driving automobile could perform well on the university campus in diverse scenarios.
{"title":"Sensor Fusion of Camera and 2D LiDAR for Self-Driving Automobile in Obstacle Avoidance Scenarios","authors":"Tan-Thien-Nien Nguyen, Thanh-Danh Phan, Minh-Thien Duong, Chi-Tam Nguyen, Hong-Phong Ly, M. Le","doi":"10.1109/IWIS56333.2022.9920917","DOIUrl":"https://doi.org/10.1109/IWIS56333.2022.9920917","url":null,"abstract":"Obstacle dodging and overtaking are the pivotal tasks ensuring safety for self-driving automobiles. Multi-sensors fusion is the must-required condition to explore the entire surrounding information. This paper proposes a novel frontal dynamic car dodging strategy for automobiles with the left-hand side steering wheel by fusion of a camera and 2D LiDAR features. To begin with, we improve the LiteSeg model to extract the segmented mask, which can determine the drivable area and the avoiding direction. In addition to a camera, 2D LiDAR is used to check the scene information on the right side, which the camera's range cannot cover. As for point clouds output of 2D LiDAR, we adopt the Adaptive Breakpoint Detection (defined as ABD) algorithm to cluster the objects in a scanning plane. Subsequently, the RANSAC algorithm forms a straight line from the clustered point clouds to determine the boundary of the right-side obstacle. Besides, we compute the distance from LiDAR to the estimated straight line to maintain a safe distance when overtaking. Last but not least, the post-processing results of the two devices are fused to decide on the obstacle dodging and overtaking. The comprehensive experiments reveal that our self-driving automobile could perform well on the university campus in diverse scenarios.","PeriodicalId":340399,"journal":{"name":"2022 International Workshop on Intelligent Systems (IWIS)","volume":"221 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2022-08-17","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"116420187","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 : 2022-08-17DOI: 10.1109/IWIS56333.2022.9920714
Adri Priadana, M. D. Putro, Changhyun Jeong, K. Jo
A gender detector has become an essential part of digital signage to support the decision to provide relevant ads for each audience. Application installed in digital signage must be capable of running on low-cost or CPU devices to minimize system costs. This study proposed a fast face gender detector (Gender-CPU) that can sprint in real-time on CPU devices implemented on digital signage. The proposed architecture contains a superficial network with attention modules (SufiaNet). This architecture only consists of three convolution layers, making it super shallow and generating a small number of parameters. In order to redeem the lack of a super shallow network, the global attention module is assigned to improve the quality of the feature map resulting from the previous convolution layers. In the experiment, the training and validation process is conducted on the UTKFace, the Adience Benchmark, and the Labeled Faces in the Wild (LFW) datasets. The SufiaNet gains competitive accuracy compared to other common and light architectures on the three datasets. Moreover, the detector can run 84.97 frames per second on a CPU device, which is fast to run in real-time.
性别探测器已经成为数字标牌的重要组成部分，以支持为每个受众提供相关广告的决定。安装在数字标牌中的应用程序必须能够在低成本或CPU设备上运行，以最大限度地降低系统成本。本研究提出了一种快速的人脸性别检测器(gender -CPU)，可以在数字标牌上实现CPU设备的实时冲刺。所提出的体系结构包含一个具有注意力模块的表层网络(SufiaNet)。该架构仅由三个卷积层组成，使其非常浅，并且生成少量参数。为了弥补超浅网络的不足，分配了全局关注模块，以提高前几层卷积得到的特征映射的质量。在实验中，在UTKFace、Adience Benchmark和Labeled Faces In the Wild (LFW)数据集上进行了训练和验证过程。在这三个数据集上，SufiaNet获得了与其他常见和轻量级架构相比具有竞争力的精度。此外，检测器可以在CPU设备上每秒运行84.97帧，实时运行速度很快。
{"title":"A Fast Real-time Face Gender Detector on CPU using Superficial Network with Attention Modules","authors":"Adri Priadana, M. D. Putro, Changhyun Jeong, K. Jo","doi":"10.1109/IWIS56333.2022.9920714","DOIUrl":"https://doi.org/10.1109/IWIS56333.2022.9920714","url":null,"abstract":"A gender detector has become an essential part of digital signage to support the decision to provide relevant ads for each audience. Application installed in digital signage must be capable of running on low-cost or CPU devices to minimize system costs. This study proposed a fast face gender detector (Gender-CPU) that can sprint in real-time on CPU devices implemented on digital signage. The proposed architecture contains a superficial network with attention modules (SufiaNet). This architecture only consists of three convolution layers, making it super shallow and generating a small number of parameters. In order to redeem the lack of a super shallow network, the global attention module is assigned to improve the quality of the feature map resulting from the previous convolution layers. In the experiment, the training and validation process is conducted on the UTKFace, the Adience Benchmark, and the Labeled Faces in the Wild (LFW) datasets. The SufiaNet gains competitive accuracy compared to other common and light architectures on the three datasets. Moreover, the detector can run 84.97 frames per second on a CPU device, which is fast to run in real-time.","PeriodicalId":340399,"journal":{"name":"2022 International Workshop on Intelligent Systems (IWIS)","volume":"50 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2022-08-17","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"128504360","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 : 2022-08-17DOI: 10.1109/IWIS56333.2022.9920759
Wendan Du, A. Abbas-Turki, A. Koukam, K. Jo
Conventional intersection managements, such as signalized intersections, may not necessarily be the optimal strategies when it comes to connected and automated vehicles (CAV) environment. Cooperative intersection management (CIM) is tailored for CAVs aiming at replacing the conventional traffic control strategies. This paper focuses on the safety constraint in the case of conflicting movements. The constraint is needed for both optimization: robot trajectory (speed profile) and sequence (which robot crosses the intersection first, which is the second one and so on). The main objective is to safely minimize the lost time between two conflicting robots while saving energy. The paper studies the safety condition of the cyber-physical system and derives the optimal control point that saves time. Quadratic optimization is used to derive the speed trajectory of the robot in order to save energy. Simulations show that the proposed approach allows safe and efficient cooperative intersection.
{"title":"Safe Cooperative Intersection of Autonomous and Connected Robots: Trajectory and Schedule Optimization","authors":"Wendan Du, A. Abbas-Turki, A. Koukam, K. Jo","doi":"10.1109/IWIS56333.2022.9920759","DOIUrl":"https://doi.org/10.1109/IWIS56333.2022.9920759","url":null,"abstract":"Conventional intersection managements, such as signalized intersections, may not necessarily be the optimal strategies when it comes to connected and automated vehicles (CAV) environment. Cooperative intersection management (CIM) is tailored for CAVs aiming at replacing the conventional traffic control strategies. This paper focuses on the safety constraint in the case of conflicting movements. The constraint is needed for both optimization: robot trajectory (speed profile) and sequence (which robot crosses the intersection first, which is the second one and so on). The main objective is to safely minimize the lost time between two conflicting robots while saving energy. The paper studies the safety condition of the cyber-physical system and derives the optimal control point that saves time. Quadratic optimization is used to derive the speed trajectory of the robot in order to save energy. Simulations show that the proposed approach allows safe and efficient cooperative intersection.","PeriodicalId":340399,"journal":{"name":"2022 International Workshop on Intelligent Systems (IWIS)","volume":"38 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2022-08-17","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"134279472","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 : 2022-08-17DOI: 10.1109/IWIS56333.2022.9920755
Jehwan Choi, Kanghyun Jo
In this paper, The network with a faster detection speed than the original YOLOv5 nano model is proposed. The network defined as a bridge module reduced the number of channels and changed the speed quickly by applying pixel-wise operation instead of using a convolution layer. Especially, element-wise addition operation of each output feature maps is the main method. As a result, the detection speed is faster than the original detection method about 30 35%. On the other hand, mAP (mean average precision) is recorded at 50.7%, which is 1.4% lower than the original detection method. However, the original detection method showed good results in 3 classes and the proposed method showed good results in 5 classes. And the proposed method detected more objects in a detection result image. Therefore, the proposed method is a more efficient object detection network.
本文提出了比原来的YOLOv5纳米模型检测速度更快的网络。定义为桥接模块的网络通过应用逐像素操作而不是使用卷积层来减少通道数量并快速改变速度。其中，各输出特征映射的逐元加法运算是主要方法。因此，检测速度比原来的检测方法快约30 - 35%。另一方面，mAP (mean average precision)为50.7%，比原检测方法降低了1.4%。然而，原检测方法在3个类别中显示出良好的结果，而本文方法在5个类别中显示出良好的结果。该方法在一幅检测结果图像中检测出更多的目标。因此，本文提出的方法是一种更高效的目标检测网络。
{"title":"Lightweight Bird Eye View Detection Network with Bridge Block Based on YOLOv5","authors":"Jehwan Choi, Kanghyun Jo","doi":"10.1109/IWIS56333.2022.9920755","DOIUrl":"https://doi.org/10.1109/IWIS56333.2022.9920755","url":null,"abstract":"In this paper, The network with a faster detection speed than the original YOLOv5 nano model is proposed. The network defined as a bridge module reduced the number of channels and changed the speed quickly by applying pixel-wise operation instead of using a convolution layer. Especially, element-wise addition operation of each output feature maps is the main method. As a result, the detection speed is faster than the original detection method about 30 35%. On the other hand, mAP (mean average precision) is recorded at 50.7%, which is 1.4% lower than the original detection method. However, the original detection method showed good results in 3 classes and the proposed method showed good results in 5 classes. And the proposed method detected more objects in a detection result image. Therefore, the proposed method is a more efficient object detection network.","PeriodicalId":340399,"journal":{"name":"2022 International Workshop on Intelligent Systems (IWIS)","volume":"1 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2022-08-17","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"117330738","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 : 2022-08-17DOI: 10.1109/IWIS56333.2022.9920868
J. Sánchez-Galán, Anel Henry Royo, K. Jo, Danilo Cáceres Hernández
This document focuses on the contributions made in the development and advances achieved in the task of automatic Feature Detection for Watermelon (Citrillus lanatus). A special interest is given to feature-based methods such as: morphological and adaptive threshold approaches, that work by extracting color and texture information. A first hand example about how these two methods can be applied to a data set comprised in export level watermelons coming from Panama is provided. Limitations of the method are discussed and a final conclusion about the field and recent avenues of work with ensemble methods is given. The importance of this document is that it helps the automatic understanding of watermelon patterns with computer vision.
{"title":"Automatic Feature Detection and Classification for Watermelon (Citrillus lanatus)","authors":"J. Sánchez-Galán, Anel Henry Royo, K. Jo, Danilo Cáceres Hernández","doi":"10.1109/IWIS56333.2022.9920868","DOIUrl":"https://doi.org/10.1109/IWIS56333.2022.9920868","url":null,"abstract":"This document focuses on the contributions made in the development and advances achieved in the task of automatic Feature Detection for Watermelon (Citrillus lanatus). A special interest is given to feature-based methods such as: morphological and adaptive threshold approaches, that work by extracting color and texture information. A first hand example about how these two methods can be applied to a data set comprised in export level watermelons coming from Panama is provided. Limitations of the method are discussed and a final conclusion about the field and recent avenues of work with ensemble methods is given. The importance of this document is that it helps the automatic understanding of watermelon patterns with computer vision.","PeriodicalId":340399,"journal":{"name":"2022 International Workshop on Intelligent Systems (IWIS)","volume":"96 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2022-08-17","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"117315190","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 : 2022-08-17DOI: 10.1109/IWIS56333.2022.9920879
Xuan-Thuy Vo, T. Tran, Duy-Linh Nguyen, K. Jo
Multiple object tracking involves multi-task learning to handle object detection and data association tasks concurrently. Conventionally, object detection consists of object classification and object localization (e.g., object regression) tasks, and data association is treated as a classification task. However, various tasks can cause inconsistent learning due to that the learning targets of object detection and data association tasks are different. Object detection focuses on positional information of objects while data association requires strong semantic information to identify same object target. Besides, advantageous character of multi-task learning is the correlation between tasks, and adopting such character in learning the networks can result in better generalization performance. However, existing multiple object tracking methods learn this information by treating multi-task branches independently. To understand the behaviours of multi-task networks in multiple object tracking, in this paper, we explore task-dependent representations through empirical experiments and observe that multi-task branches in multiple object tracking are complementary. To better learn such information, we introduce a novel Correlation Estimation (CE) module to estimate the correlation between object classification and bounding box regression based on statistical features of box regression quality. Finally, extensive experiments are conducted on the benchmark dataset MOT17. As a result, our method outperforms state-of-the-art online trackers without requiring additional training datasets.
{"title":"A Study on Efficient Multi-task Networks for Multiple Object Tracking","authors":"Xuan-Thuy Vo, T. Tran, Duy-Linh Nguyen, K. Jo","doi":"10.1109/IWIS56333.2022.9920879","DOIUrl":"https://doi.org/10.1109/IWIS56333.2022.9920879","url":null,"abstract":"Multiple object tracking involves multi-task learning to handle object detection and data association tasks concurrently. Conventionally, object detection consists of object classification and object localization (e.g., object regression) tasks, and data association is treated as a classification task. However, various tasks can cause inconsistent learning due to that the learning targets of object detection and data association tasks are different. Object detection focuses on positional information of objects while data association requires strong semantic information to identify same object target. Besides, advantageous character of multi-task learning is the correlation between tasks, and adopting such character in learning the networks can result in better generalization performance. However, existing multiple object tracking methods learn this information by treating multi-task branches independently. To understand the behaviours of multi-task networks in multiple object tracking, in this paper, we explore task-dependent representations through empirical experiments and observe that multi-task branches in multiple object tracking are complementary. To better learn such information, we introduce a novel Correlation Estimation (CE) module to estimate the correlation between object classification and bounding box regression based on statistical features of box regression quality. Finally, extensive experiments are conducted on the benchmark dataset MOT17. As a result, our method outperforms state-of-the-art online trackers without requiring additional training datasets.","PeriodicalId":340399,"journal":{"name":"2022 International Workshop on Intelligent Systems (IWIS)","volume":"7 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2022-08-17","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"116060485","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 : 2022-08-17DOI: 10.1109/IWIS56333.2022.9920838
Tomasz Kocejko, J. Rumiński, K. Jo
In this paper we present the algorithm for rapid moving object position estimation in an images acquired from pan tilt camera. Detection of a moving object in a image acquired from a moving camera might be quite challenging. Standard methods that relay on analyzing two consecutive frames are not applicable due to the changing background. To overtake this problem we decided to evaluate the possibility of calculating a disparity map based on this consecutive images. As a result we were able to obtain approximate position of moving object in real time. Using cpu the average detection time was below 0.17 second.
{"title":"Using Disparity Map for Moving Object Position Estimation in Pan Tilt Camera Images","authors":"Tomasz Kocejko, J. Rumiński, K. Jo","doi":"10.1109/IWIS56333.2022.9920838","DOIUrl":"https://doi.org/10.1109/IWIS56333.2022.9920838","url":null,"abstract":"In this paper we present the algorithm for rapid moving object position estimation in an images acquired from pan tilt camera. Detection of a moving object in a image acquired from a moving camera might be quite challenging. Standard methods that relay on analyzing two consecutive frames are not applicable due to the changing background. To overtake this problem we decided to evaluate the possibility of calculating a disparity map based on this consecutive images. As a result we were able to obtain approximate position of moving object in real time. Using cpu the average detection time was below 0.17 second.","PeriodicalId":340399,"journal":{"name":"2022 International Workshop on Intelligent Systems (IWIS)","volume":"8 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2022-08-17","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"124889885","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 : 2022-08-17DOI: 10.1109/IWIS56333.2022.9920894
Ge Cao, Qing Tang, Kanghyun Jo
This paper is employed on the unsupervised per-son re-identification (Re-ID) task which does not leverage any annotation provided by the target dataset and other datasets. Previous works have investigated the effectiveness of applying self-supervised contrastive learning, which adopts the cluster-based method to generate the pseudo label and split each cluster into multiple proxies by camera ID. This paper applies the Attribute Enhancement Module (AEM), which utilizes Graph Convolutional Network to integrate the correlations between attributes, human body parts features, and the extracted dis-criminative feature. And the experiments are implemented to demonstrate the great performance of the proposed Attribute Enhancement Contrastive Learning (AECL) in camera-agnostic version and camera-aware version on two large-scale datasets, including Market-1501 and DukeMTMC-ReID. Compared with the baseline and the state-of-the-art, the proposed framework achieves competitive results.
{"title":"Graph-based Attribute-aware Unsupervised Person Re-identification with Contrastive learning","authors":"Ge Cao, Qing Tang, Kanghyun Jo","doi":"10.1109/IWIS56333.2022.9920894","DOIUrl":"https://doi.org/10.1109/IWIS56333.2022.9920894","url":null,"abstract":"This paper is employed on the unsupervised per-son re-identification (Re-ID) task which does not leverage any annotation provided by the target dataset and other datasets. Previous works have investigated the effectiveness of applying self-supervised contrastive learning, which adopts the cluster-based method to generate the pseudo label and split each cluster into multiple proxies by camera ID. This paper applies the Attribute Enhancement Module (AEM), which utilizes Graph Convolutional Network to integrate the correlations between attributes, human body parts features, and the extracted dis-criminative feature. And the experiments are implemented to demonstrate the great performance of the proposed Attribute Enhancement Contrastive Learning (AECL) in camera-agnostic version and camera-aware version on two large-scale datasets, including Market-1501 and DukeMTMC-ReID. Compared with the baseline and the state-of-the-art, the proposed framework achieves competitive results.","PeriodicalId":340399,"journal":{"name":"2022 International Workshop on Intelligent Systems (IWIS)","volume":"8 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2022-08-17","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"132224061","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：