High-Level Semantic Feature Detection: A New Perspective for Pedestrian Detection

Liu, Wei; Liao, Shengcai; Ren, Wei; Hu, Weidong; Yan, Yu

doi:10.1109/cvpr.2019.00533

Cited by 390 publications

(399 citation statements)

References 27 publications

Supporting

Mentioning

342

Contrasting

Unclassified

Order By: Relevance

“…Although executed on the computer without GPU, the proposed method is much faster than SDN [22] executed on the computer with GPU, which has a competitive detection miss rate with our model. Very deep models such as CSP (Center and Scale Prediction) [33] back boned with ResNet-50 run on GPU obtain approximately the same speed with our model run on CPU and have much larger model parameters. Even the backbone network is changed to squeezed models such as MobileNetV1; their model is still very large and the speed-up execution time highly relies on expensive GPU hardware.…”

Section: Computational Complexity Analysismentioning

confidence: 67%

“…In recent years, many deep learning methods [28,32,33] have obtained excellent results by extracting and combining complex deep features for pedestrian detection, but the models based on deep networks have critical requirements for the hardware shown in Table 4, and TOPS (Tera Operations Per Second) are computed according to the specifications [44][45][46][47][48][49]. While the pedestrian detection is applied in an auto-driving platform, it cannot obtain huge computation resources.…”

Section: Discussionmentioning

confidence: 99%

“…Li et al [32] utilized deep feature (feature maps from ResNet-50) pyramids for multi-resolution feature extraction and realized a competitive accuracy and real-time pedestrian detection on Geforce GTX GPU. Liu et al [33] also utilized multi-resolution features from ResNet; the features are concatenated simultaneously for center point location regression and scale prediction without region proposal ahead.…”

Section: Deep Learning-based Pedestrian Detectionmentioning

confidence: 99%

“…Even the backbone network is changed to squeezed models such as MobileNetV1; their model is still very large and the speed-up execution time highly relies on expensive GPU hardware. [33] GTX 1080Ti GPU 352 0.041 12.6M CSP (ResNet-50) [33] GTX 1080Ti GPU 352 0.061 40.6M Feature Pyramid [32] GTX 1070 GPU 206 0.13 64M RPN + BF [28] Tesla K40 GPU 243 0.5 230M…”

Section: Computational Complexity Analysismentioning

confidence: 99%

See 3 more Smart Citations

Pedestrian Detection Based on Two-Stream UDN

Wang

et al. 2020

Applied Sciences

View full text Add to dashboard Cite

Pedestrian detection is the core of the driver assistance system, which collects the road conditions through the radars or cameras on the vehicle, judges whether there is a pedestrian in front of the vehicle, supports decisions such as raising the alarm, automatically slowing down, or emergency stopping to keep pedestrians safe, and improves the security when the vehicle is moving. Suffering from weather, lighting, clothing, large pose variations, and occlusion, the current pedestrian detection still has a certain distance from the practical applications. In recent years, deep networks have shown excellent performance for image detection, recognition, and classification. Some researchers employed deep network for pedestrian detection and achieve great progress, but deep networks need huge computational resources, which make it difficult to put into practical applications. In real scenarios of autonomous vehicles, the computation ability is limited. Thus, the shallow networks such as UDN (Unified Deep Networks) is a better choice, since it performs well while consuming less computation resources. Based on UDN, this paper proposes a new deep network model named two-stream UDN, which augments another branch for solving traditional UDN’s indistinction of the difference between trees/telegraph poles and pedestrians. The new branch accepts the upper third part of the pedestrian image as input, and the partial image has less deformation, stable features, and more distinguished characters from other objects. For the proposed two-stream UDN, multi-input features including the HOG (Histogram of Oriented Gradients) feature, Sobel feature, color feature, and foreground regions extracted by GrabCut segmentation algorithms are fed. Compared with the original input of UDN, the multi-input features are more conducive for pedestrian detection, since the fused HOG features and significant objects are more significant for pedestrian detection. Two-stream UDN is trained through two steps. First, the two sub-networks are trained until converge; then, we fuse results of the two subnets as the final result and feed it back to the two subnets to fine tune network parameters synchronously. To improve the performance, Swish is adopted as the activation function to obtain a faster training speed, and positive samples are mirrored and rotated with small angles to make the positive and negative samples more balanced.

show abstract

Section: Computational Complexity Analysismentioning

confidence: 67%

Section: Discussionmentioning

confidence: 99%

Section: Deep Learning-based Pedestrian Detectionmentioning

confidence: 99%

Section: Computational Complexity Analysismentioning

confidence: 99%

See 2 more Smart Citations

Pedestrian Detection Based on Two-Stream UDN

Wang

et al. 2020

Applied Sciences

View full text Add to dashboard Cite

show abstract

“…Moreover, ALF Net [4] and CSP [22] adopt other data augmentation methods including color distortion, image cropping and resizing. Similarly, we also use color distortion and random scaling in training HBAN and the results are displayed in Table 6.…”

Section: Methodsmentioning

confidence: 99%

Semantic head enhanced pedestrian detection in a crowd

Wang

2020

Neurocomputing

View full text Add to dashboard Cite

Pedestrian detection in the crowd is a challenging task because of intra-class occlusion. More prior information is needed for the detector to be robust against it. Human head area is naturally a strong cue because of its stable appearance, visibility and relative location to body. Inspired by it, we adopt an extra branch to conduct semantic head detection in parallel with traditional body branch. Instead of manually labeling the head regions, we use weak annotations inferred directly from body boxes, which is named as 'semantic head'. In this way, the head detection is formulated into using a special part of labeled box to detect the corresponding part of human body, which surprisingly improves the performance and robustness to occlusion. Moreover, the head-body alignment structure is explicitly explored by introducing Alignment Loss, which functions in a self-supervised manner. Based on these, we propose the head-body alignment net (HBAN) in this work, which aims to enhance pedestrian detection by fully utilizing the human head prior. Comprehensive evaluations are conducted to demonstrate the effectiveness of HBAN on CityPersons dataset.

show abstract

Learning based versus heuristic based: A comparative analysis of visual saliency prediction in immersive virtual reality

Aşkın

Çelikcan

2022

Computer Animation & Virtual

View full text Add to dashboard Cite

While visual saliency has been used for various purposes in virtual reality (VR), the efforts to properly understand the saliency mechanism in VR remain insufficient. In this paper, we present an extensive comparative analysis of learning-based and heuristic-based approaches to visual saliency prediction in immersive VR experienced using head-mounted-displays with a particular focus on the contribution of the depth cue. To this end, we use three learning-based RGB-D image saliency detection methods and two heuristic-based RGB-D image saliency detection methods on a VR dataset curated from three distinct virtual environments under two-dimensional and three-dimensional viewing conditions. Additionally, we extend the analysis by including a heuristic-based RGB video saliency detection method and its depth-infused version. The results acquired using these seven methods reveal the superiority of the learning-based RGB-D image saliency prediction methods in VR and validate the importance of the depth cue in the saliency prediction of virtual environments.

show abstract

High-Level Semantic Feature Detection: A New Perspective for Pedestrian Detection

Cited by 390 publications

References 27 publications

Pedestrian Detection Based on Two-Stream UDN

Pedestrian Detection Based on Two-Stream UDN

Semantic head enhanced pedestrian detection in a crowd

Learning based versus heuristic based: A comparative analysis of visual saliency prediction in immersive virtual reality

Contact Info

Product

Resources

About