A Unified Multi-Task Learning Architecture for Fast and Accurate Pedestrian Detection

Zhou, Chengju; Wu, Meiqing; Lam, Siew-Kei

doi:10.1109/tits.2020.3019390

Cited by 8 publications

(27 citation statements)

References 59 publications

(118 reference statements)

Supporting

Mentioning

Contrasting

Order By: Relevance

“…ertheless, featurizing each level of an image pyras obvious limitations. Inference time increases conly (e.g., by four times [11]), making this approach ical for real applications. Moreover, training deep…”

Section: Results Per Test Subsetmentioning

confidence: 99%

“…Our work presented in Chapter 5 that explores multi-task learning to boost detection performance for occluded pedestrians (i.e., SSAM-RCNN [11]) cannot accurately detect pedestrians at far distance, and the computational complexity are still considerably high. The extracted semantic segmentation confidence for smallscale pedestrians in RPN stage of SSAM-RCNN are inaccurate due to the high response for small-scale pedestrian proposals that are located within large-scale pedestrian regions as discussed in Section 6.1.1.2.…”

Section: Proposed Methodsmentioning

confidence: 99%

“…Proposed Method performance with downsampled input images, which significantly reduces the computational complexity. The work presented in this chapter has been published in [11].…”

Section: Pedestriansmentioning

confidence: 99%

“…In order to highlight the advantages of the proposed SS2CM RPN, we first explain the drawbacks of the recently reported multi-task learning based method SSAM-RCNN [11]. Then our proposed SS2CM RPN is introduced and its advantages over module from SSAM-RCNN are discussed.…”

Section: Semantic Segmentation To Confidence For Rpn (Ss2cm Rpn)mentioning

confidence: 99%

“…indicates that corresponding component is used. SS2CM [11] means that semantic segmentation to confidence from [11] Pedestrian detection plays an essential role in many real-world applications such as driver assistance systems, robotic navigation, pedestrian flow estimation, and video surveillance [9,19,20]. Due to its importance, pedestrian detection attracted large attention in the research community and has made significant progress [4,20].…”

mentioning

confidence: 99%

See 4 more Smart Citations

Fast and accurate vision-based pedestrian detection

Zhou¹

Self Cite

View full text Add to dashboard Cite

DateZhou Chengju sional guidance during whole period of my Ph.D. study. It is my great privilege as one of Prof. Lam Siew-Kei's students during Ph.D. study. I also wish to express my great gratitude to Dr. Wu Meiqing for her continuous guidance, professional advices and unending patience. Dr. Wu Meiqing has given me a lot of scientific discussions and technical suggestions of my thesis. Her insightful and professional suggestions are of great contributions to complete this thesis. I would like to express my sincere gratitude to the team members at Hardware and Embedded Systems Lab (HESL) during Ph.D. study. Special thanks to Mr. Chua Ngee Tat for his support during my Ph.D. study. Last but not least, I wish to thank my family for their unconditional support during my Ph.D. study. Their supports are the best encouragement for me to step forward. vi To my dear family

show abstract

Section: Results Per Test Subsetmentioning

confidence: 99%

Section: Proposed Methodsmentioning

confidence: 99%

“…Proposed Method performance with downsampled input images, which significantly reduces the computational complexity. The work presented in this chapter has been published in [11].…”

Section: Pedestriansmentioning

confidence: 99%

Section: Semantic Segmentation To Confidence For Rpn (Ss2cm Rpn)mentioning

confidence: 99%

mentioning

confidence: 99%

See 3 more Smart Citations

Fast and accurate vision-based pedestrian detection

Zhou¹

Self Cite

View full text Add to dashboard Cite

show abstract

Self-attention-guided scale-refined detector for pedestrian detection

Lin

Zhao

Zhang³

et al. 2022

Complex Intell. Syst.

View full text Add to dashboard Cite

Pedestrian detection has been researched for decades. Recently, an anchor-free method CSP is proposed to generate the pedestrian bounding box directly. When the predicted center deviates from the ground truth in the testing phase, the CSP model generates deviated pedestrian bounding box, which leads to false detection in occlusion situations. To handle this problem, we refine the scale regression branch of the CSP model to generate a more accurate prediction. The new scale regression branch outputs the distances between the center and the four edges of the pedestrian bounding box. Even if the predicted center deviates from the ground truth, an accurate bounding box can still be obtained. Moreover, we integrate a self-attention module into our model to take full advantage of the features in different depth layers. Our proposed model achieves better performance than the state-of-the-art detectors in comparison experiments on the two datasets, i.e., Citypersons and Caltech.

show abstract