Deep Learning Strong Parts for Pedestrian Detection

By Yonglong Tian, Ping Luo, Xiaogang Wang, Xiaoou Tang. IEEE International Conference on Computer Vision, 2015

515 citations recorded by the source corpus.

Recent advances in pedestrian detection are attained by transferring the learned features of Convolutional Neural Network (ConvNet) to pedestrians. This ConvNet is typically pre-trained with massive general object categories (e.g. ImageNet). Although these features are able to handle variations such as poses, viewpoints, and lightings, they may fail when pedestrian images with complex occlusions are present. Occlusion handling is one of the most important problem in pedestrian detection. Unlike previous deep models that directly learned a single detector for pedestrian detection, we propose DeepParts, which consists of extensive part detectors. DeepParts has several appealing properties. First, DeepParts can be trained on weakly labeled data, i.e. only pedestrian bounding boxes without part annotations are provided. Second, DeepParts is able to handle low IoU positive proposals that shift away from ground truth. Third, each part detector in DeepParts is a strong detector that can detect pedestrian by observing only a part of a proposal. Extensive experiments in Caltech dataset demonstrate the effectiveness of DeepParts, which yields a new state-of-the-art miss rate of 11:89%, outperforming the second best method by 10%.

Subfields: Automotive Engineering, Computer Vision and Pattern Recognition

View Paper: https://www.semanticscholar.org/paper/9d44957fea999363321ae636a1fdc21cd6780ba1

How researchers see this paper

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

In Automotive Engineering

In Computer Vision and Pattern Recognition

Citations in the corpus

Adaptive NMS: Refining Pedestrian Detection in a Crowd (2019, background): “A common framework [28, 40, 51, 7, 27, 52] for occlusion handling is to learn a series of part detectors and integrate the results to localize occluded pedestrians.”
3D Object Proposals Using Stereo Imagery for Accurate Object Class Detection (2016, background): “[44] learns part detectors with convolutional features to handle occlusion in pedestrian detec-”
Unpaired Cross-Spectral Pedestrian Detection Via Adversarial Feature Learning (2019, background): “Numerous approaches based on deep networks [3, 4, 5] have been proposed and achieved impressive performance improvement.”
Centered Weight Normalization in Accelerating Training of Deep Neural Networks (2017, background): “, image classification, speech recognition and object detection [33, 35, 14, 36, 7, 25].”
Exploit All the Layers: Fast and Accurate CNN Object Detector with Scale Dependent Pooling and Cascaded Rejection Classifiers (2016, unknown): “71 - - DeepParts [35] - - - 70.”
Learning Efficient Single-Stage Pedestrian Detectors by Asymptotic Localization Fitting (2018, methodology): “DeepParts [37] uses the LDCF detector [29] to generate proposals and then trains an ensemble of CNN for detecting different parts.”
Is Faster R-CNN Doing Well for Pedestrian Detection? (2016, methodology): “Despite the prevalent success of deeply learned features in computer vision, current leading pedestrian detectors ( e.g. , [3,4,5,6]) are in general hybrid methods that combines traditional, hand-crafted features [7,8] and deep convolutional features [9,10].”
Temporal-Context Enhanced Detection of Heavily Occluded Pedestrians (2020, unknown): “6 DeepParts [3] 22.”
Occlusion-aware R-CNN: Detecting Pedestrians in a Crowd (2018, methodology): “Several methods [35,36,49,32,72,56,46,17,16] use part-based model to describe”
Detection in Crowded Scenes: One Proposal, Multiple Predictions (2020, background): “There are other approaches on crowded detection, for example, part-based detectors [44, 38, 46, 3], which is mainly used in detecting special instances such as pedestrian.”
KGSNet: Key-Point-Guided Super-Resolution Network for Pedestrian Detection in the Wild (2020, methodology): “…reasonable, all, medium, and heavy evaluation protocols, we compare our proposed KGSNet approach with some state-of-the-art methods: DeepParts [28], F-DNN+SS [20], FasterRCNN+ATT [40], CompACT-Deep [78], SA-FastRCNN [61], RPN+BF [22], MS-CNN [79], GDFL [39], TLL-TFA [41], SDS-RCNN [19], PDOE…”
Object Detection in 20 Years: A Survey (2019, background): “idering the attraction of target and the repulsion of other surrounding objects [318]. Target occlusion is another problem that usually comes up with dense pedestrians. The ensemble of part detectors [319, 320] and the attention mechanism [321] are the most common ways to improve occluded pedestrian”
Bi-box Regression for Pedestrian Detection and Occlusion Estimation (2018, methodology): “A common framework for occlusion handling is learning and integrating a set of part detectors to handle a variety of occlusions [36, 28, 12, 11, 22, 21, 25, 23, 43, 31, 44].”
Computer vision and deep learning techniques for pedestrian detection and tracking: A survey (2018, background): “In [116] , the main idea is to construct multi-parts detectors that covers several scales of different body parts and automatically choose important parts for occlusion handling.”
CityPersons: A Diverse Dataset for Pedestrian Detection (2017, background): “Initial attempts to apply convnets for pedestrian detection, used existing detectors (mainly decision forests over hand-crafted features [2, 33]) outputs and re-scored them with a convnet classifier (plus bounding box regression) [16, 28, 1, 27].”
A Unified Multi-scale Deep Convolutional Neural Network for Fast Object Detection (2016, methodology): “8 (b) and (c) show that it performs very well for small and occluded objects, outperforming DeepParts [39], which explicitly addresses occlusion.”