Filters








1,841 Hits in 4.2 sec

Accurate Bounding-box Regression with Distance-IoU Loss for Visual Tracking [article]

Di Yuan, Nana Fan, Xiaojun Chang, Qiao Liu, Zhenyu He
2021 arXiv   pre-print
The DIoU loss can maintain the advantage provided by the IoU loss while minimizing the distance between the center points of two bounding boxes, thereby making the target estimation more accurate.  ...  the objective function when a given bounding box is completely contained within/without another bounding box; this makes it very challenging to accurately estimate the target state.  ...  All of these comparative results show that the adopted Distance-IoU loss can effectively improve the target bounding-box regression model for accurate target location and estimation.  ... 
arXiv:2007.01864v3 fatcat:x5qya4y55nd3ba6elsegh3ix5y

UAST: Uncertainty-Aware Siamese Tracking

Dawei Zhang, Yanwei Fu, Zhonglong Zheng
2022 International Conference on Machine Learning  
Recent anchor-free Siamese trackers rely on predicting the distances to four sides for efficient regression but fail to estimate accurate bounding box in complex scenes.  ...  Visual object tracking is basically formulated as target classification and bounding box estimation.  ...  L reg is the IoU Loss for bounding box regression, while L df l forces the model to focus on learning the probabilities of values neighbored with the target box, leading to a reasonable distribution.  ... 
dblp:conf/icml/0002FZ22 fatcat:dscqii6rpvfy5mr5skk6l5jnru

SiamMFC: Visual Object Tracking Based on Mainfold Full Convolution Siamese Network

Jia Chen, Fan Wang, Yingjie Zhang, Yibo Ai, Weidong Zhang
2021 Sensors  
Visual tracking task is divided into classification and regression tasks, and manifold features are introduced to improve the performance of the tracker.  ...  the target tracking performance.  ...  Conflicts of Interest: The authors declared no potential conflict of interest with respect to the research, authorship, and/or publication of this article.  ... 
doi:10.3390/s21196388 pmid:34640706 pmcid:PMC8512198 fatcat:2smd36ummje4hhq5yujvxi3fjm

SiamRCR: Reciprocal Classification and Regression for Visual Object Tracking [article]

Jinlong Peng, Zhengkai Jiang, Yueyang Gu, Yang Wu, Yabiao Wang, Ying Tai, Chengjie Wang, Weiyao Lin
2021 arXiv   pre-print
Generally, they select the bounding-box with maximum classification confidence as the final prediction.  ...  Recently, most siamese network based trackers locate targets via object classification and bounding-box regression.  ...  With localization branch, SiamRCR can predict the localization score of the regressed bounding box, making the final tracking score more consistent with the real IoU than the classification score.  ... 
arXiv:2105.11237v4 fatcat:betvik6jujbdffznmvnkwpname

Robust Template Adjustment Siamese Network for Object Visual Tracking

Chuanming Tang, Peng Qin, Jianlin Zhang
2021 Sensors  
The proposed framework TA-Siam consists of two simple subnetworks: The template adjustment subnetwork for feature extraction and the classification-regression subnetwork for bounding box prediction.  ...  For more efficient learning and faster convergence, our proposed tracker uses a more effective regression loss in the training process.  ...  Loss Function with Distance Constraint In our proposed work, the distance-IOU loss [31] is applied for regression in the training process instead of IOU loss.  ... 
doi:10.3390/s21041466 pmid:33672468 pmcid:PMC7923413 fatcat:m7iicc3ttzehhg3gagrmxxylhm

Siamese Box Adaptive Network for Visual Tracking [article]

Zedu Chen, Bineng Zhong, Guorong Li, Shengping Zhang, Rongrong Ji
2020 arXiv   pre-print
SiamBAN views the visual tracking problem as a parallel classification and regression problem, and thus directly classifies objects and regresses their bounding boxes in a unified FCN.  ...  The no-prior box design avoids hyper-parameters associated with the candidate boxes, making SiamBAN more flexible and general.  ...  We use a cross entropy and an IoU loss for classification and box regression, respectively.  ... 
arXiv:2003.06761v2 fatcat:himcovz4sbg6fmth3obljvj3ey

TLPG-Tracker: Joint Learning of Target Localization and Proposal Generation for Visual Tracking

Siyuan Li, Zhi Zhang, Ziyu Liu, Anna Wang, Linglong Qiu, Feng Du
2020 Proceedings of the Twenty-Ninth International Joint Conference on Artificial Intelligence  
Target localization and proposal generation are two essential subtasks in generic visual tracking, and it is a challenge to address both the two efficiently.  ...  Our tracker runs at over 30 FPS and sets a new state-of-the-art on five tracking benchmarks, including LaSOT, VOT2018, TrackingNet, GOT10k, OTB2015.  ...  We compare the proposed refinement method with classical bounding box regression.  ... 
doi:10.24963/ijcai.2020/99 dblp:conf/ijcai/LiZLWQD20 fatcat:r733mgvtgrh37hvoscqy6mzu5y

Robust Aleatoric Modeling for Future Vehicle Localization

Max Hudnell, True Price, Jan-Michael Frahm
2019 2019 IEEE/CVF Conference on Computer Vision and Pattern Recognition Workshops (CVPRW)  
This paper presents a novel technique for determining future bounding boxes, representing the size and location of objects -and the predictive uncertainty of both aspects -in a transit setting.  ...  An accurate prediction of an object's future localization has the potential for drastically improving critical decision making systems.  ...  IoU distributions for our network trained with an L1 loss and p = 6 (left) and a Huber loss with p = 6 (right) at t = +1.0s. The peaks at IoU = 1 are partly due to stationary test instances.  ... 
doi:10.1109/cvprw.2019.00355 dblp:conf/cvpr/HudnellPF19 fatcat:7cynmjjuobfjjkemnjbgu52pui

Siamese Box Adaptive Network for Visual Tracking

Zedu Chen, Bineng Zhong, Guorong Li, Shengping Zhang, Rongrong Ji
2020 2020 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR)  
SiamBAN views the visual tracking problem as a parallel classification and regression problem, and thus directly classifies objects and regresses their bounding boxes in a unified FCN.  ...  The no-prior box design avoids hyper-parameters associated with the candidate boxes, making SiamBAN more flexible and general.  ...  We use a cross entropy and an IoU loss for classification and box regression, respectively.  ... 
doi:10.1109/cvpr42600.2020.00670 dblp:conf/cvpr/ChenZLZJ20 fatcat:dgxy6scwjvgnrbz3g7xljz3k5q

MultiRPN-DIDNet: Multiple RPNs and Distance-IoU Discriminative Network for Real-Time UAV Target Tracking

Li Zhuo, Bin Liu, Hui Zhang, Shiyu Zhang, Jiafeng Li
2021 Remote Sensing  
Secondly, a DIDNet is designed to correct the candidate target's bounding box finely through the fusion of multi-layer features, which is trained with the Distance-IoU loss.  ...  In this paper, we propose a real-time single-target tracking method with multiple Region Proposal Networks (RPNs) and Distance-Intersection-over-Union (Distance-IoU) Discriminative Network (DIDNet), namely  ...  Label(IoU) =      0 IoU < 0.3 −1 0.3 ≤ IoU ≤ 0.6 1 IoU > 0.6 (6) For the anchor box marked as the foreground, we use Equation (7) to calculate the true value of the bounding box regression coefficient  ... 
doi:10.3390/rs13142772 fatcat:j4fpypavz5fafp4uiftghd7f24

Affine Geometrical Region CNN for Object Tracking

Yinghong Xie, Jie Shen, Chengdong Wu
2020 IEEE Access  
Then, multi-tasks loss function including the affine transformation regression is designed to refine the affine bounding box.  ...  Finally, the affine transformation NMS (Non-maximum suppression) is used to ensure the tracking bounding box having the largest IoU value.  ...  Box regression loss consists of two parts: loss of horizontal rectangle bounding box and loss of affine transformation bounding box.  ... 
doi:10.1109/access.2020.2986498 fatcat:cs3h2i76lngbngcyvbbrlsozxy

Sparse Label Assignment for Oriented Object Detection in Aerial Images

Qi Ming, Lingjuan Miao, Zhiqiang Zhou, Junjie Song, Xue Yang
2021 Remote Sensing  
Finally, the distance rotated IoU loss is proposed to eliminate the inconsistency between the training loss and the evaluation metric for better bounding box regression.  ...  Next, to accurately detect small and densely arranged objects, we use a position-sensitive feature pyramid network (PS-FPN) with a coordinate attention module to extract position-sensitive features for  ...  Distance Rotated IoU Loss for Bounding Box Regression Another thorny issue in object detection in aerial images is the inconsistency between training loss and localization accuracy.  ... 
doi:10.3390/rs13142664 fatcat:h33uht3mevb6ni7pijrowvlg6i

PGNet: Pipeline Guidance for Human Key-Point Detection

Feng Hong, Changhua Lu, Chun Liu, Ruru Liu, Weiwei Jiang, Wei Ju, Tao Wang
2020 Entropy  
We propose a novel network structure PGNet, which contains three parts: pipeline guidance strategy (PGS); Cross-Distance-IoU Loss (CIoU); and Cascaded Fusion Feature Model (CFFM).  ...  With the complexity of the network structure and the increasing amount of computation, the balance between the time of communication and the time of calculation highlights the importance.  ...  Logic operation based on bounding box regression. Alogrithm1 IoU for two axis-Aligned BBox.  ... 
doi:10.3390/e22030369 pmid:33286143 fatcat:ozdoht5gvvaflfw4xf2nhcc3nm

Fast and Furious: Real Time End-to-End 3D Detection, Tracking and Motion Forecasting with a Single Convolutional Net [article]

Wenjie Luo, Bin Yang, Raquel Urtasun
2020 arXiv   pre-print
In this paper we propose a novel deep neural network that is able to jointly reason about 3D detection, tracking and motion forecasting given data captured by a 3D sensor.  ...  Our labels contain 3D rotated bounding box as well as track id for each vehicle.  ...  In particular, for each predicted box, we first find the ground truth box with biggest overlap in terms of intersection over union (IoU).  ... 
arXiv:2012.12395v1 fatcat:hepvneh3obbbfpn7dmuroolfsu

Learning Localization-aware Target Confidence for Siamese Visual Tracking [article]

Jiahao Nie, Han Wu, Zhiwei He, Yuxiang Yang, Mingyu Gao, Zhekang Dong
2022 arXiv   pre-print
Consequently, the resulting target confidence scores, are more discriminative for the location state, allowing accurate prediction boxes tend to be predicted as high scores.  ...  Specifically, with the proposed localization-aware dynamic label (LADL) loss and localization-aware label smoothing (LALS) strategy, collaborative optimization between the classification and regression  ...  TNL2K: TNL2K [50] is a recently proposed large-scale benchmark for tracking by natural language and bounding box initialization, with more difficult video sequences.  ... 
arXiv:2204.14093v1 fatcat:l54nn2izmfdhxlfa6pkv6ac3jy
« Previous Showing results 1 — 15 out of 1,841 results