Finding and tracking people from the bottom up

By Deva Ramanan, D.A. Forsyth. 2003 IEEE Computer Society Conference on Computer Vision and Pattern Recognition, 2003. Proceedings., 2003

339 citations recorded by the source corpus.

Subfields: Artificial Intelligence, Computer Vision and Pattern Recognition

View Paper: https://www.semanticscholar.org/paper/768b9d18ebfc5ad2de18ab613d7baa0500239de8

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

Vision-based human motion analysis: An overview (2007, methodology): “Ramanan and Forsyth [81] use simple appearance-based part detectors, aided by parallel lines.”
HumanEva: Synchronized Video and Motion Capture Dataset for Evaluation of Articulated Human Motion (2006, background): “A variety of statistical [1, 2, 3, 6, 12, 36, 37, 38] as well as deterministic methods [21, 41, 35] have been developed for tracking people from single [1, 2, 7, 12, 15, 20, 21, 25, 26, 27, 29, 37] as well as multiple [3, 6, 9, 36] views.”
Efficient Human Pose Estimation from Single Depth Images (2013, background): “Ramanan and Forsyth [26] find candidate body segments as pairs of parallel lines and cluster their appearances across frames, connecting up a skeleton based on kinematic constraints.”
Tracking People by Learning Their Appearance (2007, background): “One is a bottom-up approach that looks for candidate body parts in each frame [35].”
Survey on Emotional Body Gesture Recognition (2018, methodology): “Using parallelism for clustering the appearances was proposed in [66].”
Measure Locally, Reason Globally: Occlusion-sensitive Articulated Pose Estimation (2006, background): “These patches are connected in a kinematic tree [1, 2, 4, 11, 14, 18, 20] and the pose of the person is represented by a high-dimensional state vector that includes the position and orientation of the root limb in the global image coordinate frame and the parameters of each limb relative to its pare”
Pictorial Structures for Object Recognition (2004, neutral): “fortrackingpeoplebymatchingmodelsateachframe[ 34 ]. Incontrast,mostworkon trackinghighlyarticulatedobjectssuchaspeoplereliesheavilyonmotioninformation”
Quasi-Dense Similarity Learning for Multiple Object Tracking (2020, methodology): “Contemporary MOT methods [2,4,44,45, 54] mainly follow the tracking-by-detection paradigm [36].”
Strike a pose: tracking people by finding stylized poses (2005, methodology): “Our basic representation is a tree pictorial structure that decomposes a person model into a shape model and appearance model [6, 8, 16].”
Beyond trees: common-factor models for 2D human pose recovery (2005, methodology): “The cardboard people model of [7] uses a kinematic chain, and subsequent work by [4, 6, 9, 11] uses a full kinematic tree.”
Tracking loose-limbed people (2004, methodology): “[3], Ioffe and Forsyth [10, 11, 12] and Ramanan and Forsyth [18].”
Hierarchical Gaussian process latent variable models (2007, background): “Do not distribute. for object recognition (Felzenszwalb & Huttenlocher, 2000; Io.e & Forsyth, 2001) and human pose estimation (Ramanan & Forsyth, 2003; Sigal et al., 2004; Lan & Huttenlocher, 2005).”
HumanEva: Synchronized Video and Motion Capture Dataset and Baseline Algorithm for Evaluation of Articulated Human Motion (2010, methodology): “…2004a, 2004b; Felzenszwalb and Huttenlocher 2005; Hua et al. 2005; Lan and Huttenlocher 2005; Mori 2005, Mori et al. 2004; Ramanan et al. 2005; Ramanan and Forsyth 2003; Ren et al. 2005; Ronfard et al. 2002; Sigal and Black 2006) or multiple (Balan et al. 2005; Deutscher and Reid 2005;…”
Learning shape models for monocular human pose estimation from the Microsoft Xbox Kinect (2011, methodology): “We adopt a PSM [8] of the human body, motivated by computational efficiency and its success in previous work [19, 18, 5, 1, 14].”
Action recognition from a distributed representation of pose and appearance (2011, methodology): “Pictorial structures based algorithms like that of [9, 19, 17 , 10] deal w ith the articulated nature of humans by finding body parts like limbs and torsos and constructing the overall pose using the prior knowledge of human body structure.”
3D Traffic Scene Understanding From Movable Platforms (2014, neutral): “Conclusions are drawn in Section 6.”
Guiding model search using segmentation (2005, background): “Ioffe and Forsyth [4] use a simple rectangle detector to find candidates and assemble them by sampling based on kinematic constraints.”
A Review on Video-Based Human Activity Recognition (2013, background): “A discriminative model of appearance (color) is built either by a bottom-up approach [106] that looks for candidate body parts in each frame, or by a top-down approach [107] that looks for the entire human body.”