Point-to-Set Distance Metric Learning on Deep Representations for Visual Tracking
For autonomous driving application, a car shall be able to track objects in the scene in order to estimate where and how they will move such that the tracker embedded in the car can efficiently alert the car for effective collision-avoidance. Traditional discriminative object tracking methods usuall...
Main Authors: | , , , , , |
---|---|
Format: | Other Non-Article Part of Journal/Newspaper |
Language: | unknown |
Published: |
2017
|
Subjects: | |
Online Access: | https://figshare.com/articles/journal_contribution/Point-to-Set_Distance_Metric_Learning_on_Deep_Representations_for_Visual_Tracking/10243709 |
id |
ftleicesterunfig:oai:figshare.com:article/10243709 |
---|---|
record_format |
openpolar |
spelling |
ftleicesterunfig:oai:figshare.com:article/10243709 2023-05-15T16:01:47+02:00 Point-to-Set Distance Metric Learning on Deep Representations for Visual Tracking Shengping Zhang Yuankai Qi Feng Jiang Xiangyuan Lan Pong C. Yuen Huiyu Zhou 2017-11-20T00:00:00Z https://figshare.com/articles/journal_contribution/Point-to-Set_Distance_Metric_Learning_on_Deep_Representations_for_Visual_Tracking/10243709 unknown 2381/41049 https://figshare.com/articles/journal_contribution/Point-to-Set_Distance_Metric_Learning_on_Deep_Representations_for_Visual_Tracking/10243709 All Rights Reserved Uncategorized Metric learning point to set visual tracking Text Journal contribution 2017 ftleicesterunfig 2021-11-11T19:32:34Z For autonomous driving application, a car shall be able to track objects in the scene in order to estimate where and how they will move such that the tracker embedded in the car can efficiently alert the car for effective collision-avoidance. Traditional discriminative object tracking methods usually train a binary classifier via a support vector machine (SVM) scheme to distinguish the target from its background. Despite demonstrated success, the performance of the SVM-based trackers is limited because the classification is carried out only depending on support vectors (SVs) but the target's dynamic appearance may look similar to the training samples that have not been selected as SVs, especially when the training samples are not linearly classifiable. In such cases, the tracker may drift to the background and fail to track the target eventually. To address this problem, in this paper, we propose to integrate the point-to-set/image-to-imageSet distance metric learning (DML) into visual tracking tasks and take full advantage of all the training samples when determining the best target candidate. The point-to-set DML is conducted on convolutional neural network features of the training data extracted from the starting frames. When a new frame comes, target candidates are first projected to the common subspace using the learned mapping functions, and then the candidate having the minimal distance to the target template sets is selected as the tracking result. Extensive experimental results show that even without model update the proposed method is able to achieve favorable performance on challenging image sequences compared with several state-of-the-art trackers. Other Non-Article Part of Journal/Newspaper DML University of Leicester: Figshare |
institution |
Open Polar |
collection |
University of Leicester: Figshare |
op_collection_id |
ftleicesterunfig |
language |
unknown |
topic |
Uncategorized Metric learning point to set visual tracking |
spellingShingle |
Uncategorized Metric learning point to set visual tracking Shengping Zhang Yuankai Qi Feng Jiang Xiangyuan Lan Pong C. Yuen Huiyu Zhou Point-to-Set Distance Metric Learning on Deep Representations for Visual Tracking |
topic_facet |
Uncategorized Metric learning point to set visual tracking |
description |
For autonomous driving application, a car shall be able to track objects in the scene in order to estimate where and how they will move such that the tracker embedded in the car can efficiently alert the car for effective collision-avoidance. Traditional discriminative object tracking methods usually train a binary classifier via a support vector machine (SVM) scheme to distinguish the target from its background. Despite demonstrated success, the performance of the SVM-based trackers is limited because the classification is carried out only depending on support vectors (SVs) but the target's dynamic appearance may look similar to the training samples that have not been selected as SVs, especially when the training samples are not linearly classifiable. In such cases, the tracker may drift to the background and fail to track the target eventually. To address this problem, in this paper, we propose to integrate the point-to-set/image-to-imageSet distance metric learning (DML) into visual tracking tasks and take full advantage of all the training samples when determining the best target candidate. The point-to-set DML is conducted on convolutional neural network features of the training data extracted from the starting frames. When a new frame comes, target candidates are first projected to the common subspace using the learned mapping functions, and then the candidate having the minimal distance to the target template sets is selected as the tracking result. Extensive experimental results show that even without model update the proposed method is able to achieve favorable performance on challenging image sequences compared with several state-of-the-art trackers. |
format |
Other Non-Article Part of Journal/Newspaper |
author |
Shengping Zhang Yuankai Qi Feng Jiang Xiangyuan Lan Pong C. Yuen Huiyu Zhou |
author_facet |
Shengping Zhang Yuankai Qi Feng Jiang Xiangyuan Lan Pong C. Yuen Huiyu Zhou |
author_sort |
Shengping Zhang |
title |
Point-to-Set Distance Metric Learning on Deep Representations for Visual Tracking |
title_short |
Point-to-Set Distance Metric Learning on Deep Representations for Visual Tracking |
title_full |
Point-to-Set Distance Metric Learning on Deep Representations for Visual Tracking |
title_fullStr |
Point-to-Set Distance Metric Learning on Deep Representations for Visual Tracking |
title_full_unstemmed |
Point-to-Set Distance Metric Learning on Deep Representations for Visual Tracking |
title_sort |
point-to-set distance metric learning on deep representations for visual tracking |
publishDate |
2017 |
url |
https://figshare.com/articles/journal_contribution/Point-to-Set_Distance_Metric_Learning_on_Deep_Representations_for_Visual_Tracking/10243709 |
genre |
DML |
genre_facet |
DML |
op_relation |
2381/41049 https://figshare.com/articles/journal_contribution/Point-to-Set_Distance_Metric_Learning_on_Deep_Representations_for_Visual_Tracking/10243709 |
op_rights |
All Rights Reserved |
_version_ |
1766397512788213760 |