Using k-Poselets for Detecting People and Localizing Their Keypoints

By Georgia Gkioxari, Bharath Hariharan, Ross B. Girshick, Jitendra Malik. 2014

180 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

Hypercolumns for object segmentation and fine-grained localization (2014, methodology): “A baseline system trained using our pipeline but with just the fc7 features performs significantly worse than our system, and is even worse than a HOG-based method [21].”
Towards Accurate Multi-person Pose Estimation in the Wild (2017, methodology): “stimation. For example, Pishchulin et al. [34] follow this paradigm by using PS-based pose estimation. A more robust to occlusions person detector, modelled after poselets, is used by Gkioxari et al. [19]. Further, Yang and Ramanan [45] fuse detection and pose in one model by using a PS model. The i”
Realtime Multi-person 2D Pose Estimation Using Part Affinity Fields (2016, methodology): “A common approach [28, 11, 33, 16] is to divide and conquer the problem by employing a person detector and performing single-person pose estimation for each detection.”
From captions to visual concepts and back (2014, methodology): “To calibrate the output of different detectors, we use the image level likelihood pi to compute precision on a held-out subset of the training data [13].”
Viewpoints and keypoints (2014, methodology): “[13, 12] evaluated their keypoint prediction algorithm in this setting.”
Flowing ConvNets for Human Pose Estimation in Videos (2015, neutral): “An alternative approach uses poselets [1, 13].”
Recurrent Human Pose Estimation (2016, methodology): “An alternative approach, which also included configuration constraints, was based around the poselet idea [6], [17].”
AttentionNet: Aggregating Weak Directions for Accurate Object Detection (2015, methodology): “ith an additional improvement of +3.3%. For the reﬁnement step, we simply re-use our AttentionNet for re-localization, therefore we do not need an extra model. Similar to ours, Poselets-based methods [2,13,3] are also limited to a single-class (e.g. human) object detection. Our method outperforms th”
RF-based 3D skeletons (2018, background): “2D pose estimation has achieved remarkable success recently [6, 8, 11, 13, 16, 22, 33].”
Beyond Bounding Boxes: Precise Localization of Objects in Images (2015, methodology): “A baseline system trained using our pipeline but with just the fc7 features performs significantly worse than our system, and is even worse than a HOGbased method [28].”
Actions and Attributes from Wholes and Parts (2014, background): “Similarly, poselets [1, 3, 14, 15, 20, 28] are an ensemble of models that capture parts of an object under different viewpoints and have been used for object detection, action and attribute classification and pose estimation.”
AlphaPose: Whole-Body Regional Multi-Person Pose Estimation and Tracking in Real-Time (2022, methodology): “Previous attempts approached this problem by using either a topdown framework [9], [10] or a bottom-up framework [11], [12], [13].”
Benchmarking and Error Diagnosis in Multi-instance Pose Estimation (2017, methodology): “…the performance of human pose estimation algorithms has recently improved dramatically, thanks to the development of suitable deep architectures [9, 11, 12, 17, 28, 29, 32, 33, 42, 44] and the availability of well-annotated image datasets, such as MPII Human Pose Dataset and COCO [4, 27].”
Self-Supervised Learning of Geometrically Stable Features Through Probabilistic Introspection (2018, background): “Keypoint detection has been extremely well studied for the case of humans [14, 40, 7, 1] and recent approaches have leveraged deep architectures [35, 34].”
Do Convnets Learn Correspondence? (2014, methodology): “Human pose estimation has been approached using tree structured methods to model the spatial relationships between parts [12, 13, 14], and also using poselets [15] as an intermediate step to localize human keypoints [16, 17].”
Monocular human pose estimation: A survey of deep learning-based methods (2020, methodology): “…deep learning brings a huge impact on vision-based human pose estimation, traditional 2D HPE algorithms adopt hand-craft feature extraction and sophisticated body models to obtain local representations and global pose structures (Dantone et al., 2013; Chen and Yuille, 2014; Gkioxari et al., 2014b).”
Single-Shot Multi-person 3D Pose Estimation from Monocular RGB (2017, background): “Multi-Person 2D Pose Estimation : A common approach for multi-person 2D pose estimation is to ﬁrst detect single persons and then 2D pose [44, 13, 55, 19, 40].”
R-CNNs for Pose Estimation and Action Detection (2014, methodology): “We report keypoint prediction AP for k-poselets [14].”