DeeperCut: A Deeper, Stronger, and Faster Multi-person Pose Estimation Model

By Eldar Insafutdinov, Leonid Pishchulin, Bjoern Andres, Mykhaylo Andriluka, Bernt Schiele. 2016

1,179 citations recorded by the source corpus.

The goal of this paper is to advance the state-of-the-art of articulated pose estimation in scenes with multiple people. To that end we contribute on three fronts. We propose (1) improved body part detectors that generate effective bottom-up proposals for body parts; (2) novel image-conditioned pairwise terms that allow to assemble the proposals into a variable number of consistent body part configurations; and (3) an incremental optimization strategy that explores the search space more efficiently thus leading both to better performance and significant speed-up factors. Evaluation is done on two single-person and two multi-person pose estimation benchmarks. The proposed approach significantly outperforms best known multi-person pose estimation results while demonstrating competitive performance on the task of single person pose estimation (Models and code available at http://pose.mpi-inf.mpg.de).

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, methodology): “Recently, there has been significant progress on this problem, mostly by leveraging deep Convolutional Neural Networks (CNNs) trained on large labeled datasets [45, 27, 44, 10, 33, 2, 7, 6, 20, 25, 8].”
Realtime Multi-person 2D Pose Estimation Using Part Affinity Fields (2016, methodology): “We build a two-branch network upon CPMs [38] to iteratively refine both confidence maps and PAFs with global image and spatial contexts.”
Why Should We Add Early Exits to Neural Networks? (2020, methodology): “Apart from Inception, similar approaches have been successfully applied to, among others, face recognition [61], object detection [14], image segmentation [15], pose estimation [25], saliency detection [41], metric learning [71], visual attention prediction [69], and super resolution [34, 65].”
2D/3D Pose Estimation and Action Recognition Using Multitask Deep Learning (2018, background): “The problem of human pose estimation has been intensively studied in the last years, from Pictorial Structures [2, 17, 37] to more recent CNN approaches [34, 25, 38, 20, 41, 54, 5, 51, 52, 36].”
RGB-D-based Human Motion Recognition with Deep Learning: A Survey (2017, background): “letons directly from single images or video sequences, such as DeepPose (Toshev and Szegedy, 2014), Deepercut (Insafutdinov et al., 2016) and Adversarial PoseNet (Chen et al.”
Deep High-Resolution Representation Learning for Visual Recognition (2019, background): “Other works include: light upsample process [7], [24], [92], [152], possibly with dilated convolutions used in the back-bone [63], [89], [113]; light downsample and heavy up-sample processes [141], recombinator networks [55]; improving skip connections with more or complicated convolutional units…”
ConvNets and ImageNet Beyond Accuracy: Explanations, Bias Detection, Adversarial Examples and Model Criticism (2017, background): “For example, the state of the art image segmentation [15, 20] or pose estimation models [23, 6] heavily rely on pre-trained Ima-1 genet features.”
Recurrent Human Pose Estimation (2016, background): “However, the advent of Convolutional Neural Networks (ConvNets), together with large scale training sets, has led to models that perform well in demanding scenarios with unconstrained body postures and large appearance variations [10], [18], [37], [39].”
Compositional Human Pose Regression (2017, background): “For example, a prevalent family of state-of-the-art methods [11, 6, 31, 5, 47, 18] adopt a multi-stage architecture, where the output of the previous stage is used as inputs to enhance the learning of the next stage.”
Survey on Emotional Body Gesture Recognition (2018, background): “tion has shifted towards multi-person pose estimation [74], [78], [79].”
Pose for Action - Action for Pose (2016, background): “s [18], exemplars [23] or data dependent probabilities learned by a neural network [1]. Pose estimation in videos can be improved by taking temporal information or motion cues into account [2], [10], [12], [14], [16], [45]. In [16] several pose hypotheses are generated for each video frame and a smo”
Engineering a Less Artificial Intelligence. (2019, background): “Similarly, transfer learning has proven extremely useful in the behavioral tracking of animals: using a pre-trained network and a small number of training images (z200) for fine-tuning enables the resulting network to perform very close to human-level labeling accuracy (Mathis et al., 2018; Insafutd”
LCR-Net++: Multi-Person 2D and 3D Pose Detection in Natural Images (2018, neutral): “[78] 78.”
OpenPose: Realtime Multi-Person 2D Pose Estimation Using Part Affinity Fields (2018, methodology): “The pairwise representations used in [2], which are offset vectors between every pair of body parts, are difficult to regress precisely and thus a separate logistic regression is required to convert the pairwise features into a probability score.”
LCR-Net: Localization-Classiﬁcation-Regression for Human Pose (positive): “Most similar to our approach are the methods that jointly localize humans and estimate their 2D pose [10, 11, 21].”
Neural Body Fitting: Unifying Deep Learning and Model Based Human Pose and Shape Estimation (2018, methodology): “At test-time, we estimate person height and location in the image using 2D DeeperCut keypoints [16], and center the person within a 512x512 crop such that they have a height of 440px, which roughly corresponds to the setting seen during training.”