Camera Distance-Aware Top-Down Approach for 3D Multi-Person Pose Estimation From a Single RGB Image

Moon, Gyeongsik; Chang, Ju Yong; Lee, Kyoung Mu

doi:10.1109/iccv.2019.01023

Cited by 325 publications

(390 citation statements)

References 44 publications

(144 reference statements)

Supporting

Mentioning

386

Contrasting

Order By: Relevance

“…Table 1 summarizes the quantitative results for pose estimation using the position error metric. RGB only [ 34 ] is the state-of-the-art of 3D human pose estimation using only a single RGB camera.

estimates the human pose by minimizing the cost function composed of

, and

.…”

Section: Discussionmentioning

confidence: 99%

Resolving Position Ambiguity of IMU-Based Human Pose with a Single RGB Camera

Kaichi

Maruyama

Tada

et al. 2020

Sensors

View full text Add to dashboard Cite

Human motion capture (MoCap) plays a key role in healthcare and human–robot collaboration. Some researchers have combined orientation measurements from inertial measurement units (IMUs) and positional inference from cameras to reconstruct the 3D human motion. Their works utilize multiple cameras or depth sensors to localize the human in three dimensions. Such multiple cameras are not always available in our daily life, but just a single camera attached in a smart IP devices has recently been popular. Therefore, we present a 3D pose estimation approach from IMUs and a single camera. In order to resolve the depth ambiguity of the single camera configuration and localize the global position of the subject, we present a constraint which optimizes the foot-ground contact points. The timing and 3D positions of the ground contact are calculated from the acceleration of IMUs on foot and geometric transformation of foot position detected on image, respectively. Since the results of pose estimation is greatly affected by the failure of the detection, we design the image-based constraints to handle the outliers of positional estimates. We evaluated the performance of our approach on public 3D human pose dataset. The experiments demonstrated that the proposed constraints contributed to improve the accuracy of pose estimation in single and multiple camera setting.

show abstract

estimates the human pose by minimizing the cost function composed of

, and

.…”

Section: Discussionmentioning

confidence: 99%

Resolving Position Ambiguity of IMU-Based Human Pose with a Single RGB Camera

Kaichi

Maruyama

Tada

et al. 2020

Sensors

View full text Add to dashboard Cite

show abstract

“…In [ 81 ], an insight has been provided into how the body motion and action features from views of two cameras are correlated. A similar study [ 82 ] described the 3D pose map. We will refer to their method to extend this study using multiple cameras.…”

Section: Methodsmentioning

confidence: 99%

Nonintrusive Fine-Grained Home Care Monitoring: Characterizing Quality of In-Home Postural Changes Using Bone-Based Human Sensing

Chen

Saiki

Nakamura

2020

Sensors

View full text Add to dashboard Cite

In contrast to the physical activities of able-bodied people at home, most people who require long-term specific care (e.g., bedridden patients and patients who have difficulty walking) usually show more low-intensity slow physical activities with postural changes. Although the existing devices can detect data such as heart rate and the number of steps, they have been increasing the physical burden relying on long-term wearing. The purpose of this paper is to realize a noninvasive fine-grained home care monitoring system that is sustainable for people requiring special care. In the proposed method, we present a novel technique that integrates inexpensive camera devices and bone-based human sensing technologies to characterize the quality of in-home postural changes. We realize a local process in feature data acquisition once per second, which extends from a computer browser to Raspberry Pi. Our key idea is to regard the changes of the bounding box output by standalone pose estimation models in the shape and distance as the quality of the pose conversion, body movement, and positional changes. Furthermore, we use multiple servers to realize distributed processing that uploads data to implement home monitoring as a web service. Based on the experimental results, we conveyed our findings and advice to the subject that include where the daily living habits and the irregularity of home care timings needed improvement.

show abstract

“…The network generates human pose proposals, and then classifies the generated human poses into several anchor poses, and refines the poses through regression. Moon et al [ 10 ] proposed a camera-distance-aware top-down method. Their network consists of PoseNet, which predicts root-relative 3D poses, and RootNet, which predicts the absolute 3D pose of the root joint.…”

Section: Related Workmentioning

confidence: 99%

“…Moreover, most methods estimate a root-relative 3D pose or shape, and the acquisition of pose or shape in the camera coordinate system requires ground-truth absolute depth information. Moon et al [ 10 ] recently proposed a new method to predict the absolute depth of the root joint to solve this problem and consequently estimate the 3D poses of multiple persons in the camera coordinate system. However, the method of [ 10 ] focuses on pose estimation rather than shape reconstruction, and the entire system cannot be learned in end-to-end fashion due to the separation of pose estimation and absolute depth estimation modules.…”

Section: Introductionmentioning

confidence: 99%

“…Our framework outputs a tensor of grid structure in a single-shot manner, where each grid cell contains information about the person contained therein. Moreover, we use the distance computation method proposed in [ 10 ] to predict the absolute depth of a person in the camera coordinate system. This approach requires bounding box information tightly enclosing a person, but we assume that no bounding box information is given as input considering the single-shot approach.…”

Section: Introductionmentioning

confidence: 99%

See 1 more Smart Citation

Single-Shot 3D Multi-Person Shape Reconstruction from a Single RGB Image

Kim

Chang

2020

Entropy

Self Cite

View full text Add to dashboard Cite

Although the performance of the 3D human shape reconstruction method has improved considerably in recent years, most methods focus on a single person, reconstruct a root-relative 3D shape, and rely on ground-truth information about the absolute depth to convert the reconstruction result to the camera coordinate system. In this paper, we propose an end-to-end learning-based model for single-shot, 3D, multi-person shape reconstruction in the camera coordinate system from a single RGB image. Our network produces output tensors divided into grid cells to reconstruct the 3D shapes of multiple persons in a single-shot manner, where each grid cell contains information about the subject. Moreover, our network predicts the absolute position of the root joint while reconstructing the root-relative 3D shape, which enables reconstructing the 3D shapes of multiple persons in the camera coordinate system. The proposed network can be learned in an end-to-end manner and process images at about 37 fps to perform the 3D multi-person shape reconstruction task in real time.

show abstract

Camera Distance-Aware Top-Down Approach for 3D Multi-Person Pose Estimation From a Single RGB Image

Cited by 325 publications

References 44 publications

Resolving Position Ambiguity of IMU-Based Human Pose with a Single RGB Camera

Resolving Position Ambiguity of IMU-Based Human Pose with a Single RGB Camera

Nonintrusive Fine-Grained Home Care Monitoring: Characterizing Quality of In-Home Postural Changes Using Bone-Based Human Sensing

Single-Shot 3D Multi-Person Shape Reconstruction from a Single RGB Image

Contact Info

Product

Resources

About