Self Adversarial Training for Human Pose Estimation

By Chia-Jung Chou, Jui-Ting Chien, Hwann-Tzong Chen. 2017

172 citations recorded by the source corpus.

This paper presents a deep learning based approach to the problem of human pose estimation. We employ generative adversarial networks as our learning paradigm in which we set up two stacked hourglass networks with the same architecture, one as the generator and the other as the discriminator. The generator is used as a human pose estimator after the training is done. The discriminator distinguishes ground-truth heatmaps from generated ones, and back-propagates the adversarial loss to the generator. This process enables the generator to learn plausible human body configurations and is shown to be useful for improving the prediction accuracy.

Subfields: Artificial Intelligence, 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 Artificial Intelligence

In Computer Vision and Pattern Recognition

Citations in the corpus

2D/3D Pose Estimation and Action Recognition Using Multitask Deep Learning (2018, methodology): “Generative Adversarial Networks (GANs) have been used to improve the capacity of learning structural information [13] as well as to refine the heat maps by learning more plausible predictions [15], However, detection approaches do not provide joint coordinates directly.”
Deep High-Resolution Representation Learning for Human Pose Estimation (2019, neutral): “7 7 84 3 91 2 Ning et al. [44] 98 :1 96 3 92 2 87 8 90 6 87 6 82 7 91 2 Luvizon et al. [37] 98:1 96:6 92:0 87:5 90:6 88:0 82:7 91:2 Chu et al. [14] 98:5 96:3 91:9 88:1 90:6 88:0 85:0 91:5 Chou et al. [12] 98 :2 96 8 92 2 88 0 91 3 89 1 84 9 91 8 Chen et al. [10] 98:1 96:5 92:5 88:5 90:2 89.6 86:0 91”
Deep Learning-based Human Pose Estimation: A Survey (2020, background): “[40] built an adversarial learning-based network with two stacked hourglass networks sharing the same structure as the discriminator and generator, respectively.”
Monocular human pose estimation: A survey of deep learning-based methods (2020, background): “…et al., 2016), CPM (Wei et al., 2016), PRM (Yang et al., 2017), feed forward module (Belagiannis and Zisserman, 2017), HRNet (Sun et al., 2019), GAN (Chou et al., 2017; Chen et al., 2017; Peng et al., 2018) (3) Body structure constraint : (Tompson et al., 2014; Lifshitz et al., 2016; Yang et…”
Dynamic Kernel Distillation for Efficient Pose Estimation in Videos (2019, methodology): “To synchronize the learning speed between generator and discriminator, we follow [2, 6] to update λ in Eqn.”
Toward fast and accurate human pose estimation via soft-gated skip connections (2020, methodology): “More recent methods attempt to combine the HG based architecture with attention mechanisms [8], feature pyramids [39] and adversarial training [7], [6].”
Look at Boundary: A Boundary-Aware Face Alignment Algorithm (2018, methodology): “Following [9, 10], we introduce the idea of adversarial learning by pairing the boundary effectiveness discriminator D and the boundary heatmaps estimator G.”
Counterfactual Vision-and-Language Navigation via Adversarial Path Sampling (2019, methodology): “ing has been successfully used to guide the target during model training [13, 35, 10, 27, 18, 1].”
Integral Human Pose Regression (2017, methodology): “The most widely adopted is mean square error (or L2 distance) between the predicted heat map and ground-truth heat map with a 2D Gaussian blob centered on the ground truth joint location [41, 43, 5, 29, 8, 10, 12, 4].”
Jointly Optimize Data Augmentation and Network Training: Adversarial Data Augmentation in Human Pose Estimation (2018, methodology): “Recent applications of GANs in the human pose estimation include [6] and [7].”
Simple Baselines for Human Pose Estimation and Tracking (2018, methodology): “Hourglass [4] is the dominant approach on MPII benchmark as it is the basis for all leading methods [5,6,7,8].”
Human Pose Regression by Combining Indirect Part Detection and Contextual Information (2017, unknown): “On the MPII dataset, we evaluate our method using the “Single person” challenge [1].”