MODEC: Multimodal Decomposable Models for Human Pose Estimation

By Benjamin Sapp, B. Taskar. 2013

417 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

Towards Accurate Multi-person Pose Estimation in the Wild (2017, background): “The state-of-the-art approach for single-person pose on the MPII human pose [2] and FLIC [38] datasets is the CNN model of Newell et al. [31].”
Convolutional Pose Machines (2016, methodology): “In Figure 4 we show the improvement in accuracy on the FLIC dataset [32] as the size of the receptive field on the original image is varied by varying the architecture without significantly changing the number of parameters, through a series of experimental trials on input images normalized to a siz”
DeepPose: Human Pose Estimation via Deep Neural Networks (2013, methodology): “A different approach to capture higher-order relationship is through image-dependent PS models, which can be estimated via a global classiﬁer [26, 20, 18].”
Joint Training of a Convolutional Network and a Graphical Model for Human Pose Estimation (2014, methodology): “Toshev et al. [36] show state-of-art performance on the ‘FLIC’ [29] and ‘LSP’ [18] datasets.”
Stacked Hourglass Networks for Human Pose Estimation (2016, methodology): “This is done with FLIC by centering along the x-axis according to the torsobox annotation, but we do no vertical adjustment or scale normalization.”
Recent advances in convolutional neural networks (2015, methodology): “y taking the whole image as the input and output the ﬁnal human pose. From the experiments, DeepPose can outperform the deformable part models based methods [98–100] on two widely used datasets, FLIC [101] and LSP [102]. Except this holistic manner, some recent works exploit CNN to directly learn re”
Efficient object localization using Convolutional Networks (2014, methodology): “…in part due to the success of DeepLearning architectures - specifically Convolutional Networks (ConvNets) [21, 14, 22, 5] - but also due to the availability of ever larger and more comprehensive datasets [1, 16, 20] (our model’s predictions for difficult examples from [1] are shown in Figure 1).”
Single Image 3D Interpreter Network (2016, methodology): “During the past decade, researchers have made signiﬁcant progress in estimating keypoints on humans [12,29] and other objects [13,30].”
Articulated Pose Estimation by a Graphical Model with Image Dependent Pairwise Relations (2014, methodology): “We perform our experiments on two publicly available human pose estimation benchmarks: (i) the “Leeds Sports Poses” (LSP) dataset [10], that contains 1000 training and 1000 testing images from sport activities with annotated full-body human poses; (ii) the “Frames Labeled In Cinema” (FLIC) dataset [”
P-CNN: Pose-Based CNN Features for Action Recognition (2015, positive): “of-theart approach of Yang and Ramanan [42]. Their approach is based on a deformable part model to locate positions of body joints (head, elbow, wrist...). We re-train their model on the FLIC dataset [31]. Following [8], we extract a large set of pose conﬁgurations in each frame and link them over t”
Flowing ConvNets for Human Pose Estimation in Videos (2015, methodology): “For training, we follow [6] and use all the images annotated with upper-body parts (about 4.5K) in the FLIC dataset [28].”
Fast Neural Architecture Search of Compact Semantic Segmentation Models via Auxiliary Cells (2018, methodology): “The evaluation measure is PCKh [33] with thresholds of 0.”
Unsupervised Learning using Sequential Verification for Action Recognition (2016, methodology): “Table 5: Average PCK values for the keypoints in FLIC Init. wri elb sho head Mean AUC Random Init.”
Pose Machines: Articulated Pose Estimation via Inference Machines (2014, methodology): “Recent work [11,10] has shown that part detections conditioned on the location of larger composite parts improves pose estimation performance; however, these composite parts are often constructed to form tree graph structures [16].”
ChaLearn Looking at People Challenge 2014: Dataset and Results (2014, neutral): “samples Montalbano[8] No 16 27 532 800 1 720 800 Yes Yes Yes 20 13 858 HuPBA 8K+ [7] Yes 14 124 761 8 234 Yes Yes Yes 11 235 LEEDS SPORTS[4] No 14 28 000 2 000 Yes Yes No UIUC people[10] No 14 18 186 1 299 Yes Yes No Pascal VOC[2] Yes 5 8 500 1 218 Yes Yes No BUFFY[3] No 6 4 488 748 No Yes No PARSE[”
Cascaded Pyramid Network for Multi-person Pose Estimation (2017, background): “More specifically, the classical works [1, 34, 13, 33, 8, 44, 29, 20] formulate the problem of human keypoints estimation as a tree-structured or graphical model problem and predict keypoint locations based on hand-crafted features.”
Chained Predictions Using Convolutional Neural Networks (2016, neutral): “This simplification comes at a performance cost and has been addressed in various ways: mixture model of parts [17]; mixtures of full body models [18]; higher-order spatial relationships [19]; image dependent pictorial structures [20,21,22].”
DenseReg: Fully Convolutional Dense Shape Regression In-the-Wild (2016, background): “This is explicitly represented in the form of energy terms in deformable part-based methods (DPMs) (Felzenszwalb et al, 2008) and related probabilistic graphical model (PGM) approaches to the problem of pose estimation (Andriluka et al, 2009; Sapp et al, 2010; Yang and Ramanan, 2011; Sapp and Taskar”