Histograms of oriented gradients for human detection

By Navneet Dalal, Bill Triggs. 2005

35,789 citations recorded by the source corpus.

Subfields: 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 Computer Vision and Pattern Recognition

Citations in the corpus

Object Detection via a Multi-region and Semantic Segmentation-Aware CNN Model (2015, neutral): “This success of R-CNN comes from the fact that hand-engineered features like HOG [3] or SIFT [22] are replaced with the high level object representations produced from the last layer of a CNN model.”
Object Detection Networks on Convolutional Feature Maps (2015, methodology): “The pedestrian detection method 98 in [1] proposes HOG features used with linear SVMs.”
Spatial Pyramid Pooling in Deep Convolutional Networks for Visual Recognition (2014, methodology): “In the baseline models, the pooling layer after the last convolutional layer generates 6×6 feature maps, with two 4096-d fc layers and a 1000-way softmax layer following.”
Microsoft COCO: Common Objects in Context (2014, background): “The early evolution of object recognition datasets [22], [23], [24] facilitated the direct comparison of hundreds of image recognition algorithms while simultaneously pushing the ﬁeld towards more complex problems.”
Rich Feature Hierarchies for Accurate Object Detection and Semantic Segmentation (2013, methodology): “This paper is the first to show that a CNN can lead to dramatically higher object detection performance on PASCAL VOC as compared to systems based on simpler HOG-like features.”
Do Better ImageNet Models Transfer Better? (2018, neutral): “Before the rise of convolutional neural networks, most approaches to image understanding relied on hand-engineered feature descriptors [45, 15, 3].”
Group Normalization (2018, methodology): “Classical features of SIFT [39], HOG [9], and GIST [41] are group-wise representations by design, where each group of channels is constructed by some kind of histogram.”
Measuring Robustness to Natural Distribution Shifts in Image Classification (2020, methodology): “In the context of training sets, adversarial filtering is similar to hard negative mining, which is often used to generate training data for detection models [17, 30, 69, 83].”
Learning Features by Watching Objects Move (2016, methodology): “Next, a nearest neighbor graph is computed over the superpixels in the video using location and appearance (color histograms and HOG [6]) as features.”
Feature Pyramid Networks for Object Detection (2016, methodology): “Featurized image pyramids were heavily used in the era of hand-engineered features [5, 25].”
Learning Non-maximum Suppression (2017, neutral): “Several other clustering algorithms have been explored for the task of NMS without showing consistent gains: mean-shift clustering [6, 35], agglomerative clustering [2], affinity propagation clustering [17], and heuristic variants [25].”
DeepSetNet: Predicting Sets with Deep Neural Networks (2016, background): “Traditional approaches follow the slidingwindow paradigm [44, 7, 46, 11, 2], where each possible (or rather plausible) image region is scored independently to contain a person or not.”
You Only Look Once: Unified, Real-Time Object Detection (2015, methodology): “Detection pipelines generally start by extracting a set of robust features from input images (Haar [25], SIFT [23], HOG [4], convolutional features [6]).”
Aggregated Residual Transformations for Deep Neural Networks (2016, background): “In contrast to traditional hand-designed features ( e.g ., SIFT [29] and HOG [5]), features learned by neural networks from large-scale data [33] require minimal human involvement during training, and can be transferred to a variety of recognition tasks [7, 10, 28].”
Learning to Segment Object Candidates (2015, background): “Until recently, the dominant paradigm in object detection was the sliding window framework: a classifier is applied at every object location and scale [4, 8, 32].”
Deformable part models are convolutional neural networks (2014, methodology): “Another innovation of our approach is that rather than using histograms of oriented gradients (HOG) features [4], we apply DPM-CNN to a feature pyramid that is computed by another CNN.”
Simultaneous Detection and Segmentation (2014, methodology): “The approaches above were typically built on features such as SIFT[24] or HOG[8].”