Auto-Zooming CNN-Based Framework for Real-Time Pedestrian Detection in Outdoor Surveillance Videos
One of the challenges faced by surveillance video analysis is to detect objects from the frames. For outdoor surveillance, detection of small object like pedestrian is of particular interest. This paper proposes a fast, lightweight, and auto-zooming-based framework for small pedestrian detection. An...
| Main Authors: | , , , , |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
IEEE
2019-01-01
|
| Series: | IEEE Access |
| Subjects: | |
| Online Access: | https://ieeexplore.ieee.org/document/8781786/ |
| id |
doaj-76e6f111813b4550b5a21ae5b1ed6021 |
|---|---|
| record_format |
Article |
| spelling |
doaj-76e6f111813b4550b5a21ae5b1ed60212021-04-05T17:06:40ZengIEEEIEEE Access2169-35362019-01-01710581610582610.1109/ACCESS.2019.29319158781786Auto-Zooming CNN-Based Framework for Real-Time Pedestrian Detection in Outdoor Surveillance VideosSaghir Alfasly0https://orcid.org/0000-0002-8561-660XBeibei Liu1https://orcid.org/0000-0002-3907-773XYongjian Hu2Yufei Wang3Chang-Tsun Li4School of Electronic and Information Engineering, South China University of Technology, Guangzhou, ChinaSchool of Electronic and Information Engineering, South China University of Technology, Guangzhou, ChinaSchool of Electronic and Information Engineering, South China University of Technology, Guangzhou, ChinaSino-Singapore Joint Research Institute, Guangzhou, ChinaSchool of Information Technology, Deakin University, Geelong, VIC, AustraliaOne of the challenges faced by surveillance video analysis is to detect objects from the frames. For outdoor surveillance, detection of small object like pedestrian is of particular interest. This paper proposes a fast, lightweight, and auto-zooming-based framework for small pedestrian detection. An attentive virtual auto-zooming scheme is proposed to adaptively zoom-in the input frame by splitting it into non-overlapped tiles and pay attention to the only important tiles. Without sacrificing detection performance, we have obtained a fully convolutional pedestrian detection model which can be run on low computational resources. It has been trained on an outdoor surveillance dataset and evaluated on two specially prepared testing sets of small (far) pedestrians in outdoor surveillance. We have compared our framework performance with different single-step customized pedestrian detectors as well as the two-step detector faster R-CNN. The results validate the efficiency of our framework.https://ieeexplore.ieee.org/document/8781786/Deep convolutional neural networkoutdoor surveillancereal-time pedestrian detectionsmall pedestrian objectsvirtual auto-zooming |
| collection |
DOAJ |
| language |
English |
| format |
Article |
| sources |
DOAJ |
| author |
Saghir Alfasly Beibei Liu Yongjian Hu Yufei Wang Chang-Tsun Li |
| spellingShingle |
Saghir Alfasly Beibei Liu Yongjian Hu Yufei Wang Chang-Tsun Li Auto-Zooming CNN-Based Framework for Real-Time Pedestrian Detection in Outdoor Surveillance Videos IEEE Access Deep convolutional neural network outdoor surveillance real-time pedestrian detection small pedestrian objects virtual auto-zooming |
| author_facet |
Saghir Alfasly Beibei Liu Yongjian Hu Yufei Wang Chang-Tsun Li |
| author_sort |
Saghir Alfasly |
| title |
Auto-Zooming CNN-Based Framework for Real-Time Pedestrian Detection in Outdoor Surveillance Videos |
| title_short |
Auto-Zooming CNN-Based Framework for Real-Time Pedestrian Detection in Outdoor Surveillance Videos |
| title_full |
Auto-Zooming CNN-Based Framework for Real-Time Pedestrian Detection in Outdoor Surveillance Videos |
| title_fullStr |
Auto-Zooming CNN-Based Framework for Real-Time Pedestrian Detection in Outdoor Surveillance Videos |
| title_full_unstemmed |
Auto-Zooming CNN-Based Framework for Real-Time Pedestrian Detection in Outdoor Surveillance Videos |
| title_sort |
auto-zooming cnn-based framework for real-time pedestrian detection in outdoor surveillance videos |
| publisher |
IEEE |
| series |
IEEE Access |
| issn |
2169-3536 |
| publishDate |
2019-01-01 |
| description |
One of the challenges faced by surveillance video analysis is to detect objects from the frames. For outdoor surveillance, detection of small object like pedestrian is of particular interest. This paper proposes a fast, lightweight, and auto-zooming-based framework for small pedestrian detection. An attentive virtual auto-zooming scheme is proposed to adaptively zoom-in the input frame by splitting it into non-overlapped tiles and pay attention to the only important tiles. Without sacrificing detection performance, we have obtained a fully convolutional pedestrian detection model which can be run on low computational resources. It has been trained on an outdoor surveillance dataset and evaluated on two specially prepared testing sets of small (far) pedestrians in outdoor surveillance. We have compared our framework performance with different single-step customized pedestrian detectors as well as the two-step detector faster R-CNN. The results validate the efficiency of our framework. |
| topic |
Deep convolutional neural network outdoor surveillance real-time pedestrian detection small pedestrian objects virtual auto-zooming |
| url |
https://ieeexplore.ieee.org/document/8781786/ |
| work_keys_str_mv |
AT saghiralfasly autozoomingcnnbasedframeworkforrealtimepedestriandetectioninoutdoorsurveillancevideos AT beibeiliu autozoomingcnnbasedframeworkforrealtimepedestriandetectioninoutdoorsurveillancevideos AT yongjianhu autozoomingcnnbasedframeworkforrealtimepedestriandetectioninoutdoorsurveillancevideos AT yufeiwang autozoomingcnnbasedframeworkforrealtimepedestriandetectioninoutdoorsurveillancevideos AT changtsunli autozoomingcnnbasedframeworkforrealtimepedestriandetectioninoutdoorsurveillancevideos |
| _version_ |
1721540258226503680 |