Pedestrian detection: A benchmark

By Piotr Doll'ar, Christian Wojek, Bernt Schiele, Pietro Perona. 2009

1,503 citations recorded by the source corpus.

Subfields: Artificial Intelligence, Computer Vision and Pattern Recognition, Evaluation & Benchmarks

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

In Evaluation & Benchmarks

Citations in the corpus

Soft-NMS — Improving Object Detection with One Line of Code (2017, background): “It has many practical applications in autonomous driving [6, 9], video/image indexing [28, 22], surveillance [2, 11] etc.”
Pedestrian Detection with Unsupervised Multi-stage Feature Learning (2012, background): “In [10]’s addendum, the matching criteria is modified by replacing the union of the two boxes with the minimum of the two.”
Interpretable and Fine-Grained Visual Explanations for Convolutional Neural Networks (2019, background): “Convolutional Neural Networks (CNNs) have proven to produce state-of-the-art results on a multitude of vision benchmarks, such as ImageNet [34], Caltech [12] or Cityscapes [9] which led to CNNs being used in numerous real-world systems (e.”
Deep Predictive Coding Networks for Video Prediction and Unsupervised Learning (2016, methodology): “To assess that the network had indeed learned a robust representation, we tested on the Caltech Pedestrian dataset [8], which consists of videos from a dashboard-mounted camera on a vehicle driving around Los Angeles.”
BDD100K: A Diverse Driving Dataset for Heterogeneous Multitask Learning (2018, background): “Some of the datasets focus on particular objects such as pedestrians [9, 38].”
Domain Adaptation for Visual Applications: A Comprehensive Survey (2017, methodology): “justs an ofﬂine frame-based object detector to the visual characteristic of a new video clip. The Structure-Aware Adaptive Structural SVM (SA-SSVM) [212] adapts online the deformable part-based model [213] for pedestrian detection (see Figure 26). To handle the case when no target label is available”
Feature Selective Anchor-Free Module for Single-Shot Object Detection (2019, background): “It serves as a prerequisite for various downstream vision applications such as instance segmentation [12], facial analysis [1, 39], autonomous driving cars [6, 20], and video analysis [25, 33].”
YouTube-BoundingBoxes: A Large High-Precision Human-Annotated Data Set for Object Detection in Video (2017, background): “• The Caltech Pedestrian Detection data set [10] consists of 350,000 bounding boxes of pedestrians annotated from a vehicle driving through an urban environment.”
MOT16: A Benchmark for Multi-Object Tracking (2016, background): “Related work Benchmarks and challenges. In the recent past, the computer vision community has developed centralized benchmarks for numerous tasks including object detection [14], pedestrian detection [12], 3D reconstruction [39], optical ﬂow [7], [19], visual odometry [19], singleobject short-term t”
Chained-Tracker: Chaining Paired Attentive Regression Results for End-to-End Joint Multiple-Object Detection and Tracking (2020, methodology): “ine method KDNT [3], while it is 1.5 higher than its online version POI [3]. In addition, KDNT and POI use many extra training data, including ETHZ pedestrian dataset [39], Caltech pedestrian dataset [40] and their own collected surveillance dataset [3]. While we only use the training data of MOT16.”
Active skeleton for non-rigid object detection (2009, background): “A recent benchmark study [11] shows that we are still far from being able to produce a practically useful system for detecting articulated objects, such as pedestrians.”
Repulsion Loss: Detecting Pedestrians in a Crowd (2017, methodology): “s [33] benchmark respectively in Section5.2; ﬁnally, in Section5.3, the detector with repulsion loss is compared with the state-of-the-art methods sideby-side on both CityPersons [33] and Caltech-USA [7]. 5.1. Experiment Settings Datasets. Besides the CityPersons [33] benchmark introduced in Section”
CrowdHuman: A Benchmark for Detecting Human in a Crowd (2018, background): “eshold of NMS. A lower threshold may lead to drastically drop on recall, while a higher threshold brings more false positives. Current datasets and benchmarks for human detection, such as Caltech-USA [6], KITTI [25], CityPersons [31], and “person” subset of MSCOCO [17], have contributed to a rapid p”
KGSNet: Key-Point-Guided Super-Resolution Network for Pedestrian Detection in the Wild (2020, methodology): “2) Caltech Data Set: a) Training and evaluation splits: The Caltech data set [60] is one of the most popular and challenging data sets in the task of pedestrian detection, and a total of about 10-h VGA videos are recorded from urban scenes.”
Collaging Class-specific GANs for Semantic Image Synthesis (2021, methodology): “For cityscapes, we use (6) Cityscapes extra [4] and (7) Caltech Pedestrian [8] dataset.”
Will the Pedestrian Cross? A Study on Pedestrian Path Prediction (2014, neutral): “[5] and Enzweiler and Gavrila [6].”
A survey of deep learning techniques for autonomous driving (2019, methodology): “lips) Cambridge Urban N/A Daimler pedestrian [155] Pedestrian detection, Classiﬁcation, Segmentation, Path prediction Stereo and monocular cameras 91 GB (8 clips) Amsterdam, Beijing Urban N/A Caltech [156] Tracking, Segmentation, Object detection Monocular camera 11 GB Los Angeles (USA) Urban N/A Ta”
Unbiased look at dataset bias (2011, methodology): “Fortunately, this problem can be greatly alleviated by either changing the dataset regularly (as done in PASCAL VOC 2005-2007), or withholding the test set and limiting the number of times a team can request evaluation on it (as done in PASCAL VOC 2008+ and Caltech Pedestrian benchmark [ 4 ]).”
Video-Based Human Behavior Understanding: A Survey (2013, methodology): “Later datasets [13], [70] contain lower resolution pedestrians in larger images, recorded in cluttered urban areas.”