How Far are We from Solving Pedestrian Detection?

By Shanshan Zhang, Rodrigo Benenson, Mohamed Omran, Jan Hendrik Hosang, Bernt Schiele. Computer Vision and Pattern Recognition, 2016

495 citations recorded by the source corpus.

Encouraged by the recent progress in pedestrian detection, we investigate the gap between current state-of-the-art methods and the "perfect single frame detector". We enable our analysis by creating a human baseline for pedestrian detection (over the Caltech dataset), and by manually clustering the recurrent errors of a top detector. Our results characterise both localisation and background-versusforeground errors. To address localisation errors we study the impact of training annotation noise on the detector performance, and show that we can improve even with a small portion of sanitised training data. To address background/foreground discrimination, we study convnets for pedestrian detection, and discuss which factors affect their performance. Other than our in-depth analysis, we report top performance on the Caltech dataset, and provide a new sanitised set of training and test annotations.

Subfields: Artificial Intelligence, Computer Vision and Pattern Recognition

View Paper: https://www.semanticscholar.org/paper/8f9f3ee403a0d0deab85df00e005c206abb2d6e4

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

Unsupervised Hard Example Mining from Videos for Improved Object Detection (2018, background): “Many of these object detectors have been re-purposed for other tasks such as face detection [39,29,60,15], [31,63,62,25,57,23,66] and pedestrian detection [64,14,6,7,22,30,65], achieving impressive results [24,61,12].”
Illumination-aware Faster R-CNN for Robust Multispectral Pedestrian Detection (2018, background): “Most of initial attempts [5, 6, 17, 18, 19] to apply convnets for pedestrian detection task adopted a two-stage pipeline in an R-CNN [20] style.”
Is Faster R-CNN Doing Well for Pedestrian Detection? (2016, methodology): “The evaluation metrics in Caltech-New are MR − 2 and MR − 4 , corresponding to the log-average Miss Rate on FPPI ranges of [10 − 2 , 10 0 ] and [10 − 4 , 10 0 ], following [29].”
KGSNet: Key-Point-Guided Super-Resolution Network for Pedestrian Detection in the Wild (2020, methodology): “…occluded pedestrian detection in the wild, we strictly follow the standard Caltech [76] evaluation metrics 4 Note that all experiments related to the Caltech data set are conducted on new high-quality annotations provided by [23], which corrects some inaccurate annotations unless otherwise stated.”
Person Re-identification: Past, Present and Future (2016, background): “Typically the threshold is set to 0.5, and yet Zhang et al. [136] study the difference between a “perfect single frame detector” and an automatic detector under various IoU scores.”
Perceptual Generative Adversarial Networks for Small Object Detection (2017, unknown): “The evaluation metric is log-average Miss Rate on False Positive Per Image (FPPI) in [10−2, 10] (denoted as MR following [42]).”
EuroCity Persons: A Novel Benchmark for Person Detection in Traffic Scenes (2018, unknown): “RotatedFilters [67] and Checkerboards [42].”
Pedestrian Detection: The Elephant In The Room (2020, methodology): “All experiments on Caltech [12] are conducted using new annotations provided by [41].”
High-Level Semantic Feature Detection: A New Perspective for Pedestrian Detection (2019, methodology): “Following [51, 32, 46, 28, 52], we use the training data augmented by 10 folds (42782 frames) and test on the 4024 frames in the standard test set, all experiments are conducted on the new annotations provided by [49].”
Bi-box Regression for Pedestrian Detection and Occlusion Estimation (2018, background): “Many efforts have been made to improve its performance in recent years [3, 8, 17, 6, 33, 40, 5, 39, 41, 37, 34, 4].”
Mimicking Very Efficient Network for Object Detection (2017, methodology): “On Caltech, we train our models on the new annotated 10× training data provided by [37] and select only the images that contains ground-truth bounding boxes in the training dataset which has about 9k images in total.”
CityPersons: A Diverse Dataset for Pedestrian Detection (2017, methodology): “It has been shown that pedestrian detectors benefit from well aligned training samples [32], and conversely, training with misaligned samples will hamper results.”
Adaptive NMS: Refining Pedestrian Detection in a Crowd (2019, background): “In the literature, many solutions have been presented to handle such an issue, and similar as in general object detection, the past several years have witnessed its technical development from models relying on hand-crafted features [4, 5, 11, 48] to deep learning networks [45, 46, 50, 44, 49].”