Recovering human body configurations: combining segmentation and recognition

By Greg Mori, Xiaofeng Ren, Alexei A. Efros, Jitendra Malik. Proceedings of the 2004 IEEE Computer Society Conference on Computer Vision and Pattern Recognition, 2004. CVPR 2004., 2004

645 citations recorded by the source corpus.

View Paper: https://www.semanticscholar.org/paper/a5d6d6f5d9caaba221d785f0b92d07ce2bfa3a48

How researchers see this paper

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

Citations in the corpus

Learning generative models of mid-level structure in natural images (2012, background): “, 2006), classification (Campbell et al., 1997; Mori et al., 2004), or for generating simple 3D reconstructions from single photographs (Hoiem et al.”
An expectation maximization approach to the synergy between image segmentation and object categorization (2005, background): “Models that integrate high- and low-level processes of vision have been proposed during the previous decade by researchers in the biological vision community [27, 23, 25] but only recently have such models been made practically applicable to computer vision problems [30, 26, 6, 17, 9, 13, 22].”
Recognizing human actions using multiple features (2008, methodology): “Specifically, [16, 17] utilized the geometrical models of human body parts where the action is recognized by searching for the static postures in the image that match the target action.”
Shape-based pedestrian parsing (2011, methodology): “This approach is grounded in a long line of work starting with Ren and Malik [25] which scores multiple bottom-up segmentation hypotheses [21, 20, 26, 19, 5, 18].”
Multi-task low-rank affinity pursuit for image segmentation (2011, methodology): “…cross-feature information, our MLAP method infers a unified affinity matrix by seeking the sparsity-consistent low-rank affinities from the joint decompositions of multiple feature matrices into pairs of sparse and low-rank matrices, the latter of which is the production of image feature matrix…”
Utility data annotation with Amazon Mechanical Turk (2008, background): “The basic strategy is to collect multiple annotations for every image.”
Lazy Random Walks for Superpixel Segmentation (2014, background): “Furthermore, it is more convenient and effective to compute the region based visual features by superpixels, which will provide the important benefits for the vision tasks such as object recognition [10].”
Chaotic Invariants for Human Action Recognition (2007, background): “Some leading representations are: learned geometrical models of human body parts [19, 20], space-time pattern templates [12, 16]), 3D information ([17]), shape or form features [21, 13]), interest point based representation [23], motion/optical flow patterns [24, 18] and volumetric features [22, 15]”
Vision-based human motion analysis: An overview (2007, background): “[67] first perform image segmentation based on contour, shape and appearance cues.”
Progressive search space reduction for human pose estimation (2008, methodology): “We have annotated the ground-truth pose for four frames spread roughly evenly throughout the shot, by marking each body part by one line segment [15] (figure 7a).”
Strike a pose: tracking people by finding stylized poses (2005, background): “In general, detecting people is hard because people wear different clothes and can take on many poses; this suggests a bottom-up approach using part detectors [10] [12] [14] [18] [20].”
Articulated pose estimation with flexible mixtures-of-parts (2011, methodology): “Past work has explored the use of superpixels [21], contours [27, 26, 30], foreground/background color models [23, 10], and gradient descriptors [1, 15].”
Learning Instance Occlusion for Panoptic Segmentation (2019, methodology): “Similar methods have been developed for grammar-based models [59], scene understanding [30, 17], human pose recognition [40], segmentation [4], and object detection [28].”
TurboPixels: Fast Superpixels Using Geometric Flows (2009, background): “They have been adopted primarily by those attempting to segment, classify, or label images from labeled training data [8], [9], [10], [16], [18].”
Measure Locally, Reason Globally: Occlusion-sensitive Articulated Pose Estimation (2006, methodology): “There exist a number of efficient methods for detecting body parts in an image [9, 11, 16].”
Cross-view action recognition via view knowledge transfer (2011, background): “Some leading representations include space-time pattern templates [3, 27], shape features [36, 24, 7, 19], interest point based representations [15, 26, 29, 20], learned geometrical models of the human body parts [12], and motion/optical flow patterns [1, 25, 36].”
Learning Joint Top-Down and Bottom-up Processes for 3D Visual Inference (2006, background): “The common experience [9, 22, 2, 7, 19, 28] (besides biological arguments) seems to indicate that for speed and robustness it is also useful to consider a diagnostic (discriminative) model, designed to condition on the observation,”
Superpixels and Supervoxels in an Energy Optimization Framework (2010, background): “Many methods that need regular superpixels use normalized cuts [9, 1, 2, 4, 5].”
Learning shape models for monocular human pose estimation from the Microsoft Xbox Kinect (2011, background): “For instance, the output reveals a segmentation of the image into human/background [16, 11] and potentially pixelwise segmentation of the body into different limbs [5, 17].”