Pub Date : 2021-11-16DOI: 10.1109/AVSS52988.2021.9663843
Bokyeung Lee, Hanseok Ko
Anomaly detection in the video is a challenging problem in computer vision tasks. Deep networks recently have been successfully applied and achieved competitive performance in anomaly detection. Modern deep networks employ many modules which extract important features. The anomaly detection approaches just developed network architecture and inserted additional networks to improve performance, however, these methods generally require a tremendous amount of computational load and training parameters. Because of limitations in the real world such as field equipment, mobile system, etc., reducing the number of trainable parameters and model capacity is an important issue in anomaly detection. Moreover, the method, which improves the performance of the anomaly detection algorithm, should be developed without additional trainable parameters. In this paper, we propose a sparsity injecting module which reinforces the feature representation of the existing model and presents the abnormality score function using sparsity. In experimental results, our sparsity injecting module improves the performance of state-of-the-art methods without additional trainable parameters.
{"title":"Injecting Sparsity in Anomaly Detection for Efficient Inference","authors":"Bokyeung Lee, Hanseok Ko","doi":"10.1109/AVSS52988.2021.9663843","DOIUrl":"https://doi.org/10.1109/AVSS52988.2021.9663843","url":null,"abstract":"Anomaly detection in the video is a challenging problem in computer vision tasks. Deep networks recently have been successfully applied and achieved competitive performance in anomaly detection. Modern deep networks employ many modules which extract important features. The anomaly detection approaches just developed network architecture and inserted additional networks to improve performance, however, these methods generally require a tremendous amount of computational load and training parameters. Because of limitations in the real world such as field equipment, mobile system, etc., reducing the number of trainable parameters and model capacity is an important issue in anomaly detection. Moreover, the method, which improves the performance of the anomaly detection algorithm, should be developed without additional trainable parameters. In this paper, we propose a sparsity injecting module which reinforces the feature representation of the existing model and presents the abnormality score function using sparsity. In experimental results, our sparsity injecting module improves the performance of state-of-the-art methods without additional trainable parameters.","PeriodicalId":246327,"journal":{"name":"2021 17th IEEE International Conference on Advanced Video and Signal Based Surveillance (AVSS)","volume":"19 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2021-11-16","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"125730193","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 : 2021-11-16DOI: 10.1109/AVSS52988.2021.9663814
Vishal Pani, M. Bernet, Vincent Calcagno, L. V. Oudenhove, F. Brémond
Trichogramma wasps behaviors are studied extensively due to their effectiveness as biological control agents across the globe. However, to our knowledge, the field of intra/inter-species Trichogramma behavior is yet to be explored thoroughly. To study these behaviors it is crucial to identify and track Trichogramma individuals over a long period in a lab setup. For this, we propose a robust tracking pipeline named TrichTrack. Due to the unavailability of labeled data, we train our detector using an iterative weakly supervised method. We also use a weakly supervised method to train a Re-Identification (ReID) network by leveraging noisy tracklet sampling. This enables us to distinguish Trichogramma individuals that are indistinguishable from human eyes. We also develop a two-staged tracking module that filters out the easy association to improve its efficiency. Our method outperforms existing insect trackers on most of the MOTMetrics, specifically on ID switches and fragmentations.
{"title":"TrichTrack: Multi-Object Tracking of Small-Scale Trichogramma Wasps","authors":"Vishal Pani, M. Bernet, Vincent Calcagno, L. V. Oudenhove, F. Brémond","doi":"10.1109/AVSS52988.2021.9663814","DOIUrl":"https://doi.org/10.1109/AVSS52988.2021.9663814","url":null,"abstract":"Trichogramma wasps behaviors are studied extensively due to their effectiveness as biological control agents across the globe. However, to our knowledge, the field of intra/inter-species Trichogramma behavior is yet to be explored thoroughly. To study these behaviors it is crucial to identify and track Trichogramma individuals over a long period in a lab setup. For this, we propose a robust tracking pipeline named TrichTrack. Due to the unavailability of labeled data, we train our detector using an iterative weakly supervised method. We also use a weakly supervised method to train a Re-Identification (ReID) network by leveraging noisy tracklet sampling. This enables us to distinguish Trichogramma individuals that are indistinguishable from human eyes. We also develop a two-staged tracking module that filters out the easy association to improve its efficiency. Our method outperforms existing insect trackers on most of the MOTMetrics, specifically on ID switches and fragmentations.","PeriodicalId":246327,"journal":{"name":"2021 17th IEEE International Conference on Advanced Video and Signal Based Surveillance (AVSS)","volume":"78 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2021-11-16","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"128597866","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 : 2021-11-16DOI: 10.1109/AVSS52988.2021.9663768
Quan Minh Nguyen, Bang Le Van, Can Nguyen, Anh Le, Viet Dung Nguyen
People detection in top-view, fish-eye images is challenging as people in fish-eye images often appear in arbitrary directions and are distorted differently. Due to this unique radial geometry, axis-aligned people detectors often work poorly on fish-eye frames. Recent works account for this variability by modifying existing anchor-based detectors or relying on complex pre/post-processing. Anchor-based methods spread a set of pre-defined bounding boxes on the input image, most of which are invalid. In addition to being inefficient, this approach could lead to a significant imbalance between the positive and negative anchor boxes. In this work, we propose ARPD, a single-stage anchor-free fully convolutional network to detect arbitrarily rotated people in fish-eye images. Our network uses keypoint estimation to find the center point of each object and regress the object’s other properties directly. To capture the various orientation of people in fish-eye cameras, in addition to the center and size, ARPD also predicts the angle of each bounding box. We also propose a periodic loss function that accounts for angle periodicity and relieves the difficulty of learning small-angle oscillations. Experimental results show that our method competes favorably with state-of-the-art algorithms while running significantly faster.
{"title":"ARPD: Anchor-free Rotation-aware People Detection using Topview Fisheye Camera","authors":"Quan Minh Nguyen, Bang Le Van, Can Nguyen, Anh Le, Viet Dung Nguyen","doi":"10.1109/AVSS52988.2021.9663768","DOIUrl":"https://doi.org/10.1109/AVSS52988.2021.9663768","url":null,"abstract":"People detection in top-view, fish-eye images is challenging as people in fish-eye images often appear in arbitrary directions and are distorted differently. Due to this unique radial geometry, axis-aligned people detectors often work poorly on fish-eye frames. Recent works account for this variability by modifying existing anchor-based detectors or relying on complex pre/post-processing. Anchor-based methods spread a set of pre-defined bounding boxes on the input image, most of which are invalid. In addition to being inefficient, this approach could lead to a significant imbalance between the positive and negative anchor boxes. In this work, we propose ARPD, a single-stage anchor-free fully convolutional network to detect arbitrarily rotated people in fish-eye images. Our network uses keypoint estimation to find the center point of each object and regress the object’s other properties directly. To capture the various orientation of people in fish-eye cameras, in addition to the center and size, ARPD also predicts the angle of each bounding box. We also propose a periodic loss function that accounts for angle periodicity and relieves the difficulty of learning small-angle oscillations. Experimental results show that our method competes favorably with state-of-the-art algorithms while running significantly faster.","PeriodicalId":246327,"journal":{"name":"2021 17th IEEE International Conference on Advanced Video and Signal Based Surveillance (AVSS)","volume":"363 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2021-11-16","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"115967043","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 : 2021-11-16DOI: 10.1109/AVSS52988.2021.9663810
Snehashis Majhi, Srijan Das, F. Brémond
Video anomaly detection under weak supervision is complicated due to the difficulties in identifying the anomaly and normal instances during training, hence, resulting in non-optimal margin of separation. In this paper, we propose a framework consisting of Dissimilarity Attention Module (DAM) to discriminate the anomaly instances from normal ones both at feature level and score level. In order to decide instances to be normal or anomaly, DAM takes local spatio-temporal (i.e. clips within a video) dissimilarities into account rather than the global temporal context of a video. This allows the framework to detect anomalies in real-time (i.e. online) scenarios without the need of extra window buffer time. Further more, we adopt two-variants of DAM for learning the dissimilarities between successive video clips. The proposed framework along with DAM is validated on two large scale anomaly detection datasets i.e. UCF-Crime and ShanghaiTech, outperforming the online state-of-the-art approaches by 1.5% and 3.4% respectively. The source code and models will be available at https://github.com/snehashismajhi/DAM-Anomaly-Detection
{"title":"DAM: Dissimilarity Attention Module for Weakly-supervised Video Anomaly Detection","authors":"Snehashis Majhi, Srijan Das, F. Brémond","doi":"10.1109/AVSS52988.2021.9663810","DOIUrl":"https://doi.org/10.1109/AVSS52988.2021.9663810","url":null,"abstract":"Video anomaly detection under weak supervision is complicated due to the difficulties in identifying the anomaly and normal instances during training, hence, resulting in non-optimal margin of separation. In this paper, we propose a framework consisting of Dissimilarity Attention Module (DAM) to discriminate the anomaly instances from normal ones both at feature level and score level. In order to decide instances to be normal or anomaly, DAM takes local spatio-temporal (i.e. clips within a video) dissimilarities into account rather than the global temporal context of a video. This allows the framework to detect anomalies in real-time (i.e. online) scenarios without the need of extra window buffer time. Further more, we adopt two-variants of DAM for learning the dissimilarities between successive video clips. The proposed framework along with DAM is validated on two large scale anomaly detection datasets i.e. UCF-Crime and ShanghaiTech, outperforming the online state-of-the-art approaches by 1.5% and 3.4% respectively. The source code and models will be available at https://github.com/snehashismajhi/DAM-Anomaly-Detection","PeriodicalId":246327,"journal":{"name":"2021 17th IEEE International Conference on Advanced Video and Signal Based Surveillance (AVSS)","volume":"153 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2021-11-16","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"116733548","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 : 2021-11-16DOI: 10.1109/AVSS52988.2021.9663745
Joshua Bone, Mertcan Cokbas, M. Tezcan, J. Konrad, P. Ishwar
Person re-identification using rectilinear cameras has been thoroughly researched to date. However, the topic has received little attention for fisheye cameras and the few developed methods are appearance-based. We propose a geometry-based approach to re-identification for overhead fisheye cameras with overlapping fields of view. The main idea is that a person visible in two camera views is uniquely located in the view of one camera given their height and location in the other camera’s view. We develop a height-dependent mathematical relationship between these locations using the unified spherical model for omnidirectional cameras. We also propose a new fisheye-camera calibration method and a novel automated approach to calibration-data collection. Finally, we propose four re-identification algorithms that leverage geometric constraints and demonstrate their excellent accuracy, which vastly exceeds that of a state-of-the-art appearance-based method, on a fisheye-camera dataset we collected.
{"title":"Geometry-Based Person Re-Identification in Fisheye Stereo","authors":"Joshua Bone, Mertcan Cokbas, M. Tezcan, J. Konrad, P. Ishwar","doi":"10.1109/AVSS52988.2021.9663745","DOIUrl":"https://doi.org/10.1109/AVSS52988.2021.9663745","url":null,"abstract":"Person re-identification using rectilinear cameras has been thoroughly researched to date. However, the topic has received little attention for fisheye cameras and the few developed methods are appearance-based. We propose a geometry-based approach to re-identification for overhead fisheye cameras with overlapping fields of view. The main idea is that a person visible in two camera views is uniquely located in the view of one camera given their height and location in the other camera’s view. We develop a height-dependent mathematical relationship between these locations using the unified spherical model for omnidirectional cameras. We also propose a new fisheye-camera calibration method and a novel automated approach to calibration-data collection. Finally, we propose four re-identification algorithms that leverage geometric constraints and demonstrate their excellent accuracy, which vastly exceeds that of a state-of-the-art appearance-based method, on a fisheye-camera dataset we collected.","PeriodicalId":246327,"journal":{"name":"2021 17th IEEE International Conference on Advanced Video and Signal Based Surveillance (AVSS)","volume":"37 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2021-11-16","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"114320855","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 : 2021-11-16DOI: 10.1109/AVSS52988.2021.9663765
Camille Maurice, F. Lerasle
We propose two light-weight and specialized Spatio-Temporal Graph Convolutional Networks (ST-GCNs): one for actions characterized by the motion of the human body and a novel one we especially design to recognize particular objects configurations during human actions execution. We propose a late-fusion strategy of the predictions of both graphs networks to get the most out of the two and to clear out ambiguities in the action classification. This modular approach enables us to reduce memory cost and training times. Moreover we also propose the same late fusion mechanism to further improve the performance using a Bayesian approach. We show results on 2 public datasets: CAD-120 and Watch-n-Patch. Our late-fusion mechanism yields performance gains in accuracy of respectively + 21 percentage points (pp), + 7 pp on Watch-n-Patch and CAD-120 compared to the individual graphs. Our approach outperforms most of the significant existing approaches.
{"title":"Action Recognition with Fusion of Multiple Graph Convolutional Networks","authors":"Camille Maurice, F. Lerasle","doi":"10.1109/AVSS52988.2021.9663765","DOIUrl":"https://doi.org/10.1109/AVSS52988.2021.9663765","url":null,"abstract":"We propose two light-weight and specialized Spatio-Temporal Graph Convolutional Networks (ST-GCNs): one for actions characterized by the motion of the human body and a novel one we especially design to recognize particular objects configurations during human actions execution. We propose a late-fusion strategy of the predictions of both graphs networks to get the most out of the two and to clear out ambiguities in the action classification. This modular approach enables us to reduce memory cost and training times. Moreover we also propose the same late fusion mechanism to further improve the performance using a Bayesian approach. We show results on 2 public datasets: CAD-120 and Watch-n-Patch. Our late-fusion mechanism yields performance gains in accuracy of respectively + 21 percentage points (pp), + 7 pp on Watch-n-Patch and CAD-120 compared to the individual graphs. Our approach outperforms most of the significant existing approaches.","PeriodicalId":246327,"journal":{"name":"2021 17th IEEE International Conference on Advanced Video and Signal Based Surveillance (AVSS)","volume":"4 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2021-11-16","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"127252977","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 : 2021-11-16DOI: 10.1109/AVSS52988.2021.9663831
Prayushi Mathur, A. Singh, Syed Azeemuddin, Jayram Adoni, Prasad Adireddy
The avenue of deep learning is scarcely explored in the domain of thermal imaging. Recovering a high-resolution output from images and videos is a classical problem in many computer vision applications. In this paper, we propose an optimized pipeline for a real-time video super-resolution task using thermal camera on embedded edge device. To tackle the challenges, we make contributions in the following several aspects: 1) comparative study of selected deep learning super-resolution models; 2) constructing and optimizing an end-to-end inference pipeline; 3) using cutting edge technology to integrate the whole workflow; 4) a real-time performance was achieved using less data; 5) we have also experimented the entire pipeline on our custom thermal dataset. As a consequence, the chosen model was able to achieve a real-time speed of over 29, 36 and 45 high FPS; 32.9dB/0.889, 31.86dB/0.801 and 30.94dB/0.728 PSNR/SSIM values for 2x, 3x and 4x scaling factors respectively.
{"title":"A Real-time Super-Resolution for Surveillance Thermal Cameras using optimized pipeline on Embedded Edge Device","authors":"Prayushi Mathur, A. Singh, Syed Azeemuddin, Jayram Adoni, Prasad Adireddy","doi":"10.1109/AVSS52988.2021.9663831","DOIUrl":"https://doi.org/10.1109/AVSS52988.2021.9663831","url":null,"abstract":"The avenue of deep learning is scarcely explored in the domain of thermal imaging. Recovering a high-resolution output from images and videos is a classical problem in many computer vision applications. In this paper, we propose an optimized pipeline for a real-time video super-resolution task using thermal camera on embedded edge device. To tackle the challenges, we make contributions in the following several aspects: 1) comparative study of selected deep learning super-resolution models; 2) constructing and optimizing an end-to-end inference pipeline; 3) using cutting edge technology to integrate the whole workflow; 4) a real-time performance was achieved using less data; 5) we have also experimented the entire pipeline on our custom thermal dataset. As a consequence, the chosen model was able to achieve a real-time speed of over 29, 36 and 45 high FPS; 32.9dB/0.889, 31.86dB/0.801 and 30.94dB/0.728 PSNR/SSIM values for 2x, 3x and 4x scaling factors respectively.","PeriodicalId":246327,"journal":{"name":"2021 17th IEEE International Conference on Advanced Video and Signal Based Surveillance (AVSS)","volume":"108 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2021-11-16","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"124852218","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 : 2021-11-16DOI: 10.1109/AVSS52988.2021.9663736
Dipanjan Das, Soumyadip Maity, B. Dhara
Visual SLAM (VSLAM) has shown remarkable performance in robot navigation and its practical applicability can be enriched by building a multi-robot collaboration framework called Visual collaborative SLAM (CoSLAM). CoSLAM extends the usage of SLAM for navigating in larger areas for certain applications like inspection etc. using multiple vehicles which not only saves time but also power. Visual CoSLAM framework suffers from problems like i) Robot can start from anywhere in the scene using their own VSLAM which save both time and power ii) making the framework independent of the choice of SLAM for greater applicability of different SLAMs, iii) avoiding collision with other robots by a robust merging of two noisy maps, when the visual overlap is detected. Very few works are available in the literature which addresses the above problems in a single framework in a practical sense. In this paper, we present a framework for CoSLAM using monocular cameras addressing all the above problems. Unlike existing systems which work only on ORB SLAM, our framework is truly independent of SLAMs. We propose a deep learning based algorithm to find out the visually overlapped scene required for merging two or more 3D maps. Our Map Merging is robust in presence of outliers as we compute similarity transforms using both structural information as well as camera-camera relationships and choose one based on a statistical inference. Experimental results show that our framework is robust and works well for any individual SLAM where we demonstrate our result on ORB and EdgeSLAM which are prototypical extremes methods for map merging in a CoSLAM framework.
{"title":"An Efficient And Robust Framework For Collaborative Monocular Visual Slam","authors":"Dipanjan Das, Soumyadip Maity, B. Dhara","doi":"10.1109/AVSS52988.2021.9663736","DOIUrl":"https://doi.org/10.1109/AVSS52988.2021.9663736","url":null,"abstract":"Visual SLAM (VSLAM) has shown remarkable performance in robot navigation and its practical applicability can be enriched by building a multi-robot collaboration framework called Visual collaborative SLAM (CoSLAM). CoSLAM extends the usage of SLAM for navigating in larger areas for certain applications like inspection etc. using multiple vehicles which not only saves time but also power. Visual CoSLAM framework suffers from problems like i) Robot can start from anywhere in the scene using their own VSLAM which save both time and power ii) making the framework independent of the choice of SLAM for greater applicability of different SLAMs, iii) avoiding collision with other robots by a robust merging of two noisy maps, when the visual overlap is detected. Very few works are available in the literature which addresses the above problems in a single framework in a practical sense. In this paper, we present a framework for CoSLAM using monocular cameras addressing all the above problems. Unlike existing systems which work only on ORB SLAM, our framework is truly independent of SLAMs. We propose a deep learning based algorithm to find out the visually overlapped scene required for merging two or more 3D maps. Our Map Merging is robust in presence of outliers as we compute similarity transforms using both structural information as well as camera-camera relationships and choose one based on a statistical inference. Experimental results show that our framework is robust and works well for any individual SLAM where we demonstrate our result on ORB and EdgeSLAM which are prototypical extremes methods for map merging in a CoSLAM framework.","PeriodicalId":246327,"journal":{"name":"2021 17th IEEE International Conference on Advanced Video and Signal Based Surveillance (AVSS)","volume":"21 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2021-11-16","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"122645400","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 : 2021-11-16DOI: 10.1109/AVSS52988.2021.9663782
K. Bae, Kimin Yun, Jungchan Cho, Yuseok Bae
In many visual applications, we often encounter people with irregular postures, such as lying down. Many approaches adopted two-step methods to handle a person with irregular postures: 1) person detection and 2) posture prediction based on the detected person. However, it is challenging to detect irregular postures because the existing detectors were trained with datasets consisting of most upright postures. Therefore, we propose a new Irregular Human Posture (IHP) dataset to handle various postures captured from real-world surveillance cameras. The IHP dataset provides sufficient annotations to understand the posture of person, including segmentation, keypoints, and postural states. This paper also provides two baseline net-works for postural state estimation of the people trained on the IHP dataset. Moreover, we show that our baseline networks effectively detect the people with irregular postures that may be in an urgent situation in a surveillance environment.
{"title":"The Dataset and Baseline Models to Detect Human Postural States Robustly against Irregular Postures","authors":"K. Bae, Kimin Yun, Jungchan Cho, Yuseok Bae","doi":"10.1109/AVSS52988.2021.9663782","DOIUrl":"https://doi.org/10.1109/AVSS52988.2021.9663782","url":null,"abstract":"In many visual applications, we often encounter people with irregular postures, such as lying down. Many approaches adopted two-step methods to handle a person with irregular postures: 1) person detection and 2) posture prediction based on the detected person. However, it is challenging to detect irregular postures because the existing detectors were trained with datasets consisting of most upright postures. Therefore, we propose a new Irregular Human Posture (IHP) dataset to handle various postures captured from real-world surveillance cameras. The IHP dataset provides sufficient annotations to understand the posture of person, including segmentation, keypoints, and postural states. This paper also provides two baseline net-works for postural state estimation of the people trained on the IHP dataset. Moreover, we show that our baseline networks effectively detect the people with irregular postures that may be in an urgent situation in a surveillance environment.","PeriodicalId":246327,"journal":{"name":"2021 17th IEEE International Conference on Advanced Video and Signal Based Surveillance (AVSS)","volume":"8 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2021-11-16","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"124062887","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 : 2021-11-16DOI: 10.1109/AVSS52988.2021.9663829
Daniel Stadler, J. Beyerer
One of the biggest challenges in multi-pedestrian tracking arises in crowds, where missing detections can lead to wrong track-detection assignments, especially under heavy occlusion. In order to identify such situations, we cluster tracks and detections based on their overlaps and introduce different cluster states depending on the number of detections and tracks in a cluster. On the basis of this strategy, we make the following contributions. First, we propose a cluster-aware non-maximum suppression (CA-NMS) that leverages temporal information from tracks applying an increased IoU threshold in clusters with severe occlusion to reduce the number of missed detections, while at the same time limiting the number of duplicate detections. Second, for clusters with very high overlaps where detections are missing even with the CA-NMS, we utilize past track information to correct wrong assignments when missed targets are re-detected after occlusion. Furthermore, we propose a new tracking pipeline that combines the paradigms of tracking-by-detection and regression-based tracking to improve the association performance in crowded scenes. Putting all together, our tracker achieves competitive results w.r.t. the state-of-the-art on three multi-pedestrian tracking benchmarks. Our framework is analyzed with extensive ablative experiments and the impact of the proposed tracking components on the performance is evaluated.
{"title":"Multi-Pedestrian Tracking with Clusters","authors":"Daniel Stadler, J. Beyerer","doi":"10.1109/AVSS52988.2021.9663829","DOIUrl":"https://doi.org/10.1109/AVSS52988.2021.9663829","url":null,"abstract":"One of the biggest challenges in multi-pedestrian tracking arises in crowds, where missing detections can lead to wrong track-detection assignments, especially under heavy occlusion. In order to identify such situations, we cluster tracks and detections based on their overlaps and introduce different cluster states depending on the number of detections and tracks in a cluster. On the basis of this strategy, we make the following contributions. First, we propose a cluster-aware non-maximum suppression (CA-NMS) that leverages temporal information from tracks applying an increased IoU threshold in clusters with severe occlusion to reduce the number of missed detections, while at the same time limiting the number of duplicate detections. Second, for clusters with very high overlaps where detections are missing even with the CA-NMS, we utilize past track information to correct wrong assignments when missed targets are re-detected after occlusion. Furthermore, we propose a new tracking pipeline that combines the paradigms of tracking-by-detection and regression-based tracking to improve the association performance in crowded scenes. Putting all together, our tracker achieves competitive results w.r.t. the state-of-the-art on three multi-pedestrian tracking benchmarks. Our framework is analyzed with extensive ablative experiments and the impact of the proposed tracking components on the performance is evaluated.","PeriodicalId":246327,"journal":{"name":"2021 17th IEEE International Conference on Advanced Video and Signal Based Surveillance (AVSS)","volume":"135 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2021-11-16","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"116648325","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：