{"title":"SAR Image Denoising in High Dynamic Range with Speckle and Thermal Noise Refinement Modeling","authors":"Ji-Hoon Han, Woo-Jeoung Nam, Seong-Whan Lee","doi":"10.1109/AVSS56176.2022.9959170","DOIUrl":"https://doi.org/10.1109/AVSS56176.2022.9959170","url":null,"abstract":"Synthetic Aperture Radar (SAR) images inevitably contain speckle noise. Despeckling SAR images are typically represented as linear forms due to the consistency of denoising network training/inference settings on a linear scale. However, this leads to controversial problems when the linear SAR images are seen with the naked eye: i) restriction of representation for dark areas and ii) excessive expression for bright areas due to the high-intensity range. To overcome these problems, we propose a denoising framework that simultaneously eliminates speckle and thermal noise in the decibel (dB) domain through novel noise modeling. Our noise modeling allows the network to learn in the dB domain of the desired dynamic range, enabling stable end-to-end learning without separate spatial transformations. Our modeling is the first attempt to consider thermal noise. Experimental results show that our method is superior in quantitative and visual performance compared to the existing methods.","PeriodicalId":408581,"journal":{"name":"2022 18th IEEE International Conference on Advanced Video and Signal Based Surveillance (AVSS)","volume":"2 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2022-11-29","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"130378124","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}
K. Lee, Bhavin Jawade, D. Mohan, S. Setlur, V. Govindaraju
{"title":"Attribute De-biased Vision Transformer (AD-ViT) for Long-Term Person Re-identification","authors":"K. Lee, Bhavin Jawade, D. Mohan, S. Setlur, V. Govindaraju","doi":"10.1109/AVSS56176.2022.9959509","DOIUrl":"https://doi.org/10.1109/AVSS56176.2022.9959509","url":null,"abstract":"Person re-identification (re-ID) aims to retrieve images of the same identity from a gallery of person images across cameras and viewpoints. However, most works in person re-ID assume a short-term setting characterized by invariance in appearance. In contrast, a high visual variance can be frequently seen in a long-term setting due to changes in apparel and accessories, which makes the task more challenging. Therefore, learning identity-specific features agnostic of temporally variant features is crucial for robust long-term person Re-ID. To this end, we propose an Attribute De-biased Vision Transformer (AD-ViT) to provide direct supervision to learn identity-specific features. Specifically, we produce attribute labels for person instances and utilize them to guide our model to focus on identity features through gradient reversal. Our experiments on two long-term re-ID datasets - LTCC and NKUP show that the proposed work consistently outperforms current state-of-the-art methods.","PeriodicalId":408581,"journal":{"name":"2022 18th IEEE International Conference on Advanced Video and Signal Based Surveillance (AVSS)","volume":"19 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2022-11-29","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"133349734","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}
{"title":"Customer Behavior Recognition Adaptable for Changing Targets in Retail Environments","authors":"Jiahao Wen, Toru Abe, T. Suganuma","doi":"10.1109/AVSS56176.2022.9959409","DOIUrl":"https://doi.org/10.1109/AVSS56176.2022.9959409","url":null,"abstract":"Recognizing Customer Behavior (CB) from videos of instore cameras is important to smart retail solutions. Because of possible changes in retail needs and environments, a high degree of adaptability to different target CBs is required for Customer Behavior Recognition (CBR) methods. Existing CBR methods are mainly machine learning based models due to their remarkable recognition accuracy. However, trained models are not reusable for different target CBs. Consequently, existing CBR methods are hard to adapt to different target CBs because the necessary recollecting data and retraining models. In this paper, we propose a CBR method that recognizes CBs by combinations of primitives, each of which represents an object’s motion or objects’ relationship. Since primitives can be reused in combinations for various CBs, the proposed method is easily adaptable to changed target CBs. Experiments on two datasets indicate the good adaptability and sufficient recognition accuracy of our method.","PeriodicalId":408581,"journal":{"name":"2022 18th IEEE International Conference on Advanced Video and Signal Based Surveillance (AVSS)","volume":"9 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2022-11-29","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"133494672","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}
{"title":"Robust Camera Calibration with a Single Image by Planar Homography","authors":"Jae-Yeong Lee","doi":"10.1109/AVSS56176.2022.9959461","DOIUrl":"https://doi.org/10.1109/AVSS56176.2022.9959461","url":null,"abstract":"In surveillance applications, there are usually dozens of cameras at one site, and quick and easy calibration has primary importance. This paper presents a practical and robust camera calibration method that works with a single image of a planar marker and less sensitive to measurement noise. The method first detects a planar object on the ground plane and estimates a homography between the ground plane and the image plane. Two orthogonal non-infinite vanishing points are then estimated directly from the homography, not from the marker nor scene structure. The proposed method minimizes possible calibration error by choosing the direction of the two vanishing points to be as much different from the direction of the vanishing line as possible. Experimental results on synthetic images show that the proposed homography-based method greatly reduces calibration error in the presence of measurement noise, demonstrating the effectiveness of the method.","PeriodicalId":408581,"journal":{"name":"2022 18th IEEE International Conference on Advanced Video and Signal Based Surveillance (AVSS)","volume":"1 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2022-11-29","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"129850496","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}
{"title":"Distance-dependent Feature Alignment and Selection for Imbalance 3D Point Cloud Object Detection","authors":"Ming-Jen Chang, Chih-Jen Cheng, Ching-Chun Hsiao, I-Fan Chou, Ching-Chun Huang","doi":"10.1109/AVSS56176.2022.9959572","DOIUrl":"https://doi.org/10.1109/AVSS56176.2022.9959572","url":null,"abstract":"Although pillar-based 3D object detection methods can balance the performance and inference speed, the inconsistent object features caused by dramatic sparsity drops of LiDAR point clouds sabotage the detection accuracy. We present a novel and efficient plug-in method, SVDnet, to improve the state-of-the-art pillar-based models. First, a novel low-rank objective loss is introduced to extract distance-aware vehicle features and suppress the other variations. Next, we alleviated the remaining feature inconsistency caused by object positions with two strategies. One is a Distance Alignment Ratio-generation Network (DARN), which fuses multi-scale features by distance-adaptive ratios. The other is a position attention network that modulates features based on positions. Our results on the KITTI dataset show that SVDnet improves the pillar methods and outperforms the other plug-in strategies in accuracy and speed.","PeriodicalId":408581,"journal":{"name":"2022 18th IEEE International Conference on Advanced Video and Signal Based Surveillance (AVSS)","volume":"43 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2022-11-29","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"132851714","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}
{"title":"A Visual Surveillance System to Observe Realistic Road User Behavior for Improved Pedestrian and Cyclist Safety at Crossroads","authors":"Nadezda Kirillova, Horst Possegger, H. Bischof","doi":"10.1109/AVSS56176.2022.9959562","DOIUrl":"https://doi.org/10.1109/AVSS56176.2022.9959562","url":null,"abstract":"Pedestrians and cyclists suffer the most serious injuries in traffic accidents. Existing Pedestrian Protection Systems and Road Safety Systems rely on an ideal model of pedestrian behavior and do not consider that people tend to take shortcuts, appear at unexpected places or can be distracted on the road, for example, by using a smartphone or wearing headphones. Collecting and analyzing realistic road user behavior is a crucial component to improve pedestrian and cyclist safety. However, such real-world data is still missing. To address this, we propose a visual surveillance system with two perpendicular partially overlapping fields of view, combined with a fully automated deep learning-based pipeline to process and collect video observations, detect and extract road user trajectories in real-world coordinates and estimate human attributes, such as age, gender, smartphone usage, etc. We demonstrate our prototype by deploying it in two locations in a European city.","PeriodicalId":408581,"journal":{"name":"2022 18th IEEE International Conference on Advanced Video and Signal Based Surveillance (AVSS)","volume":"1 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2022-11-29","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"130962883","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}
R. Bhatt, Simrandeep Singh, Priyanka Choudhary, Mukesh Saini
{"title":"An Experimental Study of the Concept Drift Challenge in Farm Intrusion Detection using Audio","authors":"R. Bhatt, Simrandeep Singh, Priyanka Choudhary, Mukesh Saini","doi":"10.1109/AVSS56176.2022.9959493","DOIUrl":"https://doi.org/10.1109/AVSS56176.2022.9959493","url":null,"abstract":"Intrusion detection in farm settings is a challenging task. The data distribution suffers significant drift due to variations in environmental sounds. In this paper, we study the effect of this drift on state-of-the-art deep models. We experimentally found that the traditional models fail to deal with such variations and exhibit performance degradation. VGG16 turns out to be the best deep model, which shows an improvement of 2.76% over the best-performing state-of-the-art deep model on parameter F1-score. Consequently, we make an initial attempt to overcome this drift by integrating an unsupervised background noise component with standard models. On denoised signals, we obtained an average improvement of 50% on parameter accuracy for VGG16. The experimental analysis demonstrates the need for adaptive learning to handle the Spatio-temporal drifts in outdoor farm settings.","PeriodicalId":408581,"journal":{"name":"2022 18th IEEE International Conference on Advanced Video and Signal Based Surveillance (AVSS)","volume":"47 5 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2022-11-29","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"115943340","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}
Seong-Yun Jeong, Ho–Joong Kim, Myeong-Seok Oh, Gun-Hee Lee, Seonghyeon Lee
{"title":"Temporal-Invariant Video Representation Learning with Dynamic Temporal Resolutions","authors":"Seong-Yun Jeong, Ho–Joong Kim, Myeong-Seok Oh, Gun-Hee Lee, Seonghyeon Lee","doi":"10.1109/AVSS56176.2022.9959310","DOIUrl":"https://doi.org/10.1109/AVSS56176.2022.9959310","url":null,"abstract":"Recent studies for similarity-based self-supervised representation learning tend to consider only fixed temporal coverage from a given video. However, this approach limits that a model learns temporally persistent representations since it cannot reflect spatial and temporal information gaps from resolution variations. To overcome the limitation, this paper proposes a Temporal Adaptive Teacher-Student (TATS) framework that encourages the trained model to be robust on spatio-temporal variations. Our key approach is optimizing similarity-based learning that utilizes several views with dynamic temporal resolutions. From a given video, TATS captures spatio-temporal invariant clues for temporally persistent representation with cross-resolution correspondence between local and global views. Extensive experiments show that our TATS achieves competitive downstream (action recognition and video retrieval) performances on benchmarks (UCF101 and HMDB51).","PeriodicalId":408581,"journal":{"name":"2022 18th IEEE International Conference on Advanced Video and Signal Based Surveillance (AVSS)","volume":"11 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2022-11-29","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"134530711","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}
{"title":"RCRL: Replay-based Continual Representation Learning in Multi-task Super-Resolution","authors":"Jin-Yeol Park, Minhan Kim, Simon S. Woo","doi":"10.1109/AVSS56176.2022.9959552","DOIUrl":"https://doi.org/10.1109/AVSS56176.2022.9959552","url":null,"abstract":"Super-resolution (SR) aims to recover high-resolution (HR) images from low-resolution (LR) images. Recently, various attempts, e.g., unsupervised SR models and domain-specific SR have achieved outstanding performance for various real-world applications. However, they significantly suffer from low generalization performance when trained on another domain dataset. Furthermore, they often exhibit performance degradation when the model continually learns multiple tasks; so-called catastrophic forgetting degrades the SR performance. In this paper, we are the first to propose a novel approach for continual multi-task SR named Replay-based Continual Representation Learning framework that can be applicable to GAN-based SR models, which utilizes feature memory for preserving the learned features from the previous task. Our experimental results demonstrate the effectiveness of RCRL in continual multi-task SR at improving generalization performance and alleviating catastrophic forgetting.","PeriodicalId":408581,"journal":{"name":"2022 18th IEEE International Conference on Advanced Video and Signal Based Surveillance (AVSS)","volume":"47 10","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2022-11-29","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"131873694","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}
{"title":"3D Object Detection of Cars and Pedestrians by Deep Neural Networks from Unit-Sharing One-Shot NAS","authors":"O. Chen, Yu-Xuan Chang, Yu-Wei Jhao, Chih-Yu Chung, Yun-Ling Chang, Weihua Huang","doi":"10.1109/AVSS56176.2022.9959427","DOIUrl":"https://doi.org/10.1109/AVSS56176.2022.9959427","url":null,"abstract":"This work adopts camera, lidar and radar sensors to realize 3D object detection of cars and humans in varied and complicated road scenes. Based on the Radar Voxel Fusion Net (RVF-Net) consisting of voxel feature extraction module, feature aggregation module, and detection head module, we extend layers of the voxel feature extraction module with dimensions at a consistent scaling ratio, and increase a prediction head in the detection head module. According to this modified RVF-Net, the depth of the voxel feature extraction module is searched by the proposed unit-sharing one-shot Neural Architecture Search (NAS) scheme where there are five choices of bypass, and one to four copy units for each layer. Only four copy units with conditional links yield five choices for structure searches. Such an approach can minimize the search cost and parameter amount. By using the forward-looking sensing data of nuScenes, the network from the proposed NAS, and the modified RVF-Net can yield superior Average Precisions (AP), more than 2.3% AP increase, than the conventional ones in detecting cars and pedestrians simultaneously. Therefore, the proposed network can effectively accomplish multi-object recognition tasks for self-driving applications.","PeriodicalId":408581,"journal":{"name":"2022 18th IEEE International Conference on Advanced Video and Signal Based Surveillance (AVSS)","volume":"20 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2022-11-29","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"115429077","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}