Joint Deep Learning for Pedestrian Detection

By Wanli Ouyang, Xiaogang Wang. 2013

672 citations recorded by the source corpus.

Feature extraction, deformation handling, occlusion handling, and classification are four important components in pedestrian detection. Existing methods learn or design these components either individually or sequentially. The interaction among these components is not yet well explored. This paper proposes that they should be jointly learned in order to maximize their strengths through cooperation. We formulate these four components into a joint deep learning framework and propose a new deep network architecture. By establishing automatic, mutual interaction among components, the deep model achieves a 9% reduction in the average miss rate compared with the current best-performing pedestrian detection approaches on the largest Caltech benchmark dataset.

Subfields: Artificial Intelligence, Computer Vision and Pattern Recognition

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Artificial Intelligence

In Computer Vision and Pattern Recognition

Citations in the corpus

ImageNet Large Scale Visual Recognition Challenge (2014, background): “7 The Chinese University of Hong Kong Wanli Ouyang, Ping Luo, Xingyu Zeng, Shi Qiu, Yonglong Tian, Hongsheng Li, Shuo Yang, Zhe Wang, Yuanjun Xiong, Chen Qian, Zhenyao Zhu, Ruohui Wang, Chen-Change Loy, Xiaogang Wang, Xiaoou Tang (Ouyang et al., 2014; Ouyang and Wang, 2013) DeepCNet 17.”
Deformable part models are convolutional neural networks (2014, methodology): “Our work is most closely related to the deep pedestrian detection model of Ouyang and Wang [25].”
Image Super-Resolution Using Deep Convolutional Networks (2014, background): “First, its structure is intentionally designed with simplicity in mind, and yet provides superior accuracy2 compared with state-of-the-art example-based methods.”
HCP: A Flexible CNN Framework for Multi-Label Image Classification (2014, background): “Recently, in contrast to the hand-crafted features, learnt image features with deep network structures have shown their great potential in various vision recognition tasks [26], [21], [24], [36].”
DeepID-Net: multi-stage and deformable deep convolutional neural networks for object detection (2014, methodology): “In short, defpooling is max-pooling with deformation constraint. Example 2. Suppose V = k y, H = k x, i= 1,··· ,V, and j = 1,··· ,H, then the def-pooling layer degenerates to the deformation layer in [38]. There is only one output for M in this case. The deformation layer can represent the widely us”
Compositional Human Pose Regression (2017, neutral): “he graphical models have been integrated into deep networks in various ways. Tompson et al. [46] ﬁrstly combine a convolutional network with a graphical model for human pose estimation. Ouyang et al. [32] joints feature extraction, part deformation handling, occlusion handling and classiﬁcation all ”
Learning Deep Representation with Large-Scale Attributes (2015, background): “Deep learning is effective for many vision tasks [17, 34, 43, 14, 19, 11, 26, 28, 23, 25, 44, 47, 21, 49, 5].”
PersonNet: Person Re-identification with Deep Convolutional Neural Networks (2016, background): “Convolutional Neural Networks (CNNs) have proven highly successful at image recognition problems and various surveillance applications including pedestrian detection [18], [19], and tracking [20].”
Learning Complexity-Aware Cascades for Deep Pedestrian Detection (2015, background): “Recently, [15] transferred the RCNN framework to the pedestrian detection task, showing some improvement over previous deep learning detectors [27, 22].”
Pansharpening by Convolutional Neural Networks (2016, methodology): “Hence, the first layer computes 64 feature maps using a 9 × 9 receptive field and a ReLU (rectified linear unit, max(0, x) [40]) nonlinearity.”
Real-Time Object Detection With Reduced Region Proposal Network via Multi-Feature Concatenation (2020, background): “Prune [1–5] in the green region denotes pruning in convolution layers while prune [6 and 7] in the blue region represents pruning in FC layers.”
DeepReID: Deep Filter Pairing Neural Network for Person Re-identification (2014, background): “Although some deep learning works [32, 39] share the spirit of jointly optimizing components of vision systems, their problems, challenges, models and training strategies are completely different from ours.”
Deeply Learning Deformable Facial Action Parts Model for Dynamic Expression Analysis (2014, methodology): “Similar to [7] and [12], we conduct part-wise discriminative learning by partial connection to layer P6.”
Person Re-identification in the Wild (2016, methodology): “Previous works in pedestrian detection [10, 28, 43] usually use Average Precision or LogAverage Miss Rate under IoU > 0.”
Real-Time Pedestrian Detection with Deep Network Cascades (2015, methodology): “Some methods focus on improving the base features used [6, 9, 10, 42], whereas others focus on the learning algorithms [13, 25], or other techniques such as incorporating Deformable Parts Models [33, 40] or using context [13, 29, 40, 41].”