The original paper is in English. Non-English content has been machine-translated and may contain typographical errors or mistranslations. ex. Some numerals are expressed as "XNUMX".
Copyrights notice
The original paper is in English. Non-English content has been machine-translated and may contain typographical errors or mistranslations. Copyrights notice
자연적인 교통 상황에서는 가려진 부분이 많기 때문에 차량 감지가 어렵습니다. 폐색으로 인해 감지기의 훈련 전략으로 인해 특징과 레이블이 일치하지 않을 수 있습니다. 결과적으로 일부 예측 경계 상자가 주변 차량으로 이동하여 신뢰도가 낮아질 수 있습니다. 이러한 경계 상자는 AP 값을 낮추게 됩니다. 이 편지에서 우리는 이 문제를 해결하기 위한 새로운 접근 방식을 제안합니다. 도로정보를 기반으로 현재 차량의 보이는 부분의 중심을 계산합니다. 그런 다음 가변 반경 가우스 가중치 기반 방법을 적용하여 SSD 학습 시간에 보이는 부분의 중심을 기준으로 손실 함수에서 각 앵커 상자의 가중치를 다시 적용합니다. 재가중 방법은 더 높은 신뢰도와 더 정확한 경계 상자를 예측하는 기능을 갖추고 있습니다. 게다가 이 모델은 속도도 빠르고 엔드투엔드(end-to-end) 학습이 가능합니다. 실험 결과는 우리가 제안한 방법이 속도와 정확성 측면에서 일부 경쟁 방법보다 우수하다는 것을 보여줍니다.
Yu HUANG
South China University of Technology
Zhiheng ZHOU
South China University of Technology
Tianlei WANG
Wuyi University
Qian CAO
South China University of Technology
Junchu HUANG
South China University of Technology
Zirong CHEN
South China University of Technology
The copyright of the original papers published on this site belongs to IEICE. Unauthorized use of the original or translated papers is prohibited. See IEICE Provisions on Copyright for details.
부
Yu HUANG, Zhiheng ZHOU, Tianlei WANG, Qian CAO, Junchu HUANG, Zirong CHEN, "A Part-Based Gaussian Reweighted Approach for Occluded Vehicle Detection" in IEICE TRANSACTIONS on Information,
vol. E102-D, no. 5, pp. 1097-1101, May 2019, doi: 10.1587/transinf.2018EDL8257.
Abstract: Vehicle detection is challenging in natural traffic scenes because there exist a lot of occlusion. Because of occlusion, detector's training strategy may lead to mismatch between features and labels. As a result, some predicted bounding boxes may shift to surrounding vehicles and lead to lower confidences. These bounding boxes will lead to lower AP value. In this letter, we propose a new approach to address this problem. We calculate the center of visible part of current vehicle based on road information. Then a variable-radius Gaussian weight based method is applied to reweight each anchor box in loss function based on the center of visible part in training time of SSD. The reweighted method has ability to predict higher confidences and more accurate bounding boxes. Besides, the model also has high speed and can be trained end-to-end. Experimental results show that our proposed method outperforms some competitive methods in terms of speed and accuracy.
URL: https://global.ieice.org/en_transactions/information/10.1587/transinf.2018EDL8257/_p
부
@ARTICLE{e102-d_5_1097,
author={Yu HUANG, Zhiheng ZHOU, Tianlei WANG, Qian CAO, Junchu HUANG, Zirong CHEN, },
journal={IEICE TRANSACTIONS on Information},
title={A Part-Based Gaussian Reweighted Approach for Occluded Vehicle Detection},
year={2019},
volume={E102-D},
number={5},
pages={1097-1101},
abstract={Vehicle detection is challenging in natural traffic scenes because there exist a lot of occlusion. Because of occlusion, detector's training strategy may lead to mismatch between features and labels. As a result, some predicted bounding boxes may shift to surrounding vehicles and lead to lower confidences. These bounding boxes will lead to lower AP value. In this letter, we propose a new approach to address this problem. We calculate the center of visible part of current vehicle based on road information. Then a variable-radius Gaussian weight based method is applied to reweight each anchor box in loss function based on the center of visible part in training time of SSD. The reweighted method has ability to predict higher confidences and more accurate bounding boxes. Besides, the model also has high speed and can be trained end-to-end. Experimental results show that our proposed method outperforms some competitive methods in terms of speed and accuracy.},
keywords={},
doi={10.1587/transinf.2018EDL8257},
ISSN={1745-1361},
month={May},}
부
TY - JOUR
TI - A Part-Based Gaussian Reweighted Approach for Occluded Vehicle Detection
T2 - IEICE TRANSACTIONS on Information
SP - 1097
EP - 1101
AU - Yu HUANG
AU - Zhiheng ZHOU
AU - Tianlei WANG
AU - Qian CAO
AU - Junchu HUANG
AU - Zirong CHEN
PY - 2019
DO - 10.1587/transinf.2018EDL8257
JO - IEICE TRANSACTIONS on Information
SN - 1745-1361
VL - E102-D
IS - 5
JA - IEICE TRANSACTIONS on Information
Y1 - May 2019
AB - Vehicle detection is challenging in natural traffic scenes because there exist a lot of occlusion. Because of occlusion, detector's training strategy may lead to mismatch between features and labels. As a result, some predicted bounding boxes may shift to surrounding vehicles and lead to lower confidences. These bounding boxes will lead to lower AP value. In this letter, we propose a new approach to address this problem. We calculate the center of visible part of current vehicle based on road information. Then a variable-radius Gaussian weight based method is applied to reweight each anchor box in loss function based on the center of visible part in training time of SSD. The reweighted method has ability to predict higher confidences and more accurate bounding boxes. Besides, the model also has high speed and can be trained end-to-end. Experimental results show that our proposed method outperforms some competitive methods in terms of speed and accuracy.
ER -