Bo Wang
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Previous Organizations / Institutes
The organizations / institutes from the scholar's other articles in the past ten years
Bo Wang has a total of 70 co-authors. Here are the top co-authors by number of shared publications:
- Guido Gerig
New York University
10 shared publications - Marcel Prastawa
GE Global Research
10 shared publications - John Darrell Van Horn
Department of Psychology, University of Virginia
8 shared publications Paul M. Vespa
Brain Injury Research Center, Department of Neurosurgery, David Geffen School of Medicine, University of California, Los Angeles, CA 90095 USA
7 shared publicationsDong C. Liu
Sichuan University
5 shared publications- Andrei Irimia
Ethel Percy Andrus Gerontology Center, Leonard Davis School of Gerontology, University of Southern California, Los Angeles, USA
4 shared publications
Latest Publications
- 11 September 2026
- Published by Springer Nature
- p. 386-404
Token-Based Affordance Grounding with Large Vision-Language Models
- 11 September 2026
- Published by Springer Nature
- p. 386-404
Token-Based Affordance Grounding with Large Vision-Language Models
- 23 November 2024
- Published by Springer Nature
- p. 221-239
Domain-Adaptive 2D Human Pose Estimation via Dual Teachers in Extremely Low-Light Conditions
- 23 November 2024
- Published by Springer Nature
- p. 221-239
Domain-Adaptive 2D Human Pose Estimation via Dual Teachers in Extremely Low-Light Conditions
- Open Access
Domain-Adaptive 2D Human Pose Estimation via Dual Teachers in Extremely Low-Light Conditions
Existing 2D human pose estimation research predominantly concentrates on well-lit scenarios, with limited exploration of poor lighting conditions, which are a prevalent aspect of daily life. Recent studies on low-light pose estimation require the use of paired well-lit and low-light images with ground truths for training, which are impractical due to the inherent challenges associated with annotation on low-light images. To this end, we introduce a novel approach that eliminates the need for low-light ground truths. Our primary novelty lies in leveraging two complementary-teacher networks to generate more reliable pseudo labels, enabling our model achieves competitive performance on extremely low-light images without the need for training with low-light ground truths. Our framework consists of two stages. In the first stage, our model is trained on well-lit data with low-light augmentations. In the second stage, we propose a dual-teacher framework to utilize the unlabeled low-light data, where a center-based main teacher produces the pseudo labels for relatively visible cases, while a keypoints-based complementary teacher focuses on producing the pseudo labels for the missed persons of the main teacher. With the pseudo labels from both teachers, we propose a person-specific low-light augmentation to challenge a student model in training to outperform the teachers. Experimental results on real low-light dataset (ExLPose-OCN) show, our method achieves 6.8% (2.4 AP) improvement over the state-of-the-art (SOTA) method, despite no low-light ground-truth data is used in our approach, in contrast to the SOTA method. Our code will be available at:https://github.com/ayh015-dev/DA-LLPose.
Open AccessDomain-Adaptive 2D Human Pose Estimation via Dual Teachers in Extremely Low-Light Conditions
Existing 2D human pose estimation research predominantly concentrates on well-lit scenarios, with limited exploration of poor lighting conditions, which are a prevalent aspect of daily life. Recent studies on low-light pose estimation require the use of paired well-lit and low-light images with ground truths for training, which are impractical due to the inherent challenges associated with annotation on low-light images. To this end, we introduce a novel approach that eliminates the need for low-light ground truths. Our primary novelty lies in leveraging two complementary-teacher networks to generate more reliable pseudo labels, enabling our model achieves competitive performance on extremely low-light images without the need for training with low-light ground truths. Our framework consists of two stages. In the first stage, our model is trained on well-lit data with low-light augmentations. In the second stage, we propose a dual-teacher framework to utilize the unlabeled low-light data, where a center-based main teacher produces the pseudo labels for relatively visible cases, while a keypoints-based complementary teacher focuses on producing the pseudo labels for the missed persons of the main teacher. With the pseudo labels from both teachers, we propose a person-specific low-light augmentation to challenge a student model in training to outperform the teachers. Experimental results on real low-light dataset (ExLPose-OCN) show, our method achieves 6.8% (2.4 AP) improvement over the state-of-the-art (SOTA) method, despite no low-light ground-truth data is used in our approach, in contrast to the SOTA method. Our code will be available at:https://github.com/ayh015-dev/DA-LLPose.
- 26 April 2022
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Pattern Analysis and Machine Intelligence
- Vol. 45 (2) , 1636-1651
Dual Networks Based 3D Multi-Person Pose Estimation From Monocular Video
Monocular 3D human pose estimation has made progress in recent years. Most of the methods focus on single persons, which estimate the poses in the person-centric coordinates, i.e., the coordinates based on the center of the target person. Hence, these methods are inapplicable for multi-person 3D pose estimation, where the absolute coordinates (e.g., the camera coordinates) are required. Moreover, multi-person pose estimation is more challenging than single pose estimation, due to inter-person occlusion and close human interactions. Existing top-down multi-person methods rely on human detection (i.e., top-down approach), and thus suffer from the detection errors and cannot produce reliable pose estimation in multi-person scenes. Meanwhile, existing bottom-up methods that do not use human detection are not affected by detection errors, but since they process all persons in a scene at once, they are prone to errors, particularly for persons in small scales. To address all these challenges, we propose the integration of top-down and bottom-up approaches to exploit their strengths. Our top-down network estimates human joints from all persons instead of one in an image patch, making it robust to possible erroneous bounding boxes. Our bottom-up network incorporates human-detection based normalized heatmaps, allowing the network to be more robust in handling scale variations. Finally, the estimated 3D poses from the top-down and bottom-up networks are fed into our integration network for final 3D poses. To address the common gaps between training and testing data, we do optimization during the test time, by refining the estimated 3D human poses using high-order temporal constraint, re-projection loss, and bone length regularizations. We also introduce a two-person pose discriminator that enforces natural two-person interactions. Finally, we apply a semi-supervised method to overcome the 3D ground-truth data scarcity. Our evaluations demonstrate the effectiveness of the proposed method and its individual components. Our code and pretrained models are available publicly: https://github.com/3dpose/3D-Multi-Person-Pose .
- 26 April 2022
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Pattern Analysis and Machine Intelligence
- Vol. 45 (2) , 1636-1651
Dual Networks Based 3D Multi-Person Pose Estimation From Monocular Video
Monocular 3D human pose estimation has made progress in recent years. Most of the methods focus on single persons, which estimate the poses in the person-centric coordinates, i.e., the coordinates based on the center of the target person. Hence, these methods are inapplicable for multi-person 3D pose estimation, where the absolute coordinates (e.g., the camera coordinates) are required. Moreover, multi-person pose estimation is more challenging than single pose estimation, due to inter-person occlusion and close human interactions. Existing top-down multi-person methods rely on human detection (i.e., top-down approach), and thus suffer from the detection errors and cannot produce reliable pose estimation in multi-person scenes. Meanwhile, existing bottom-up methods that do not use human detection are not affected by detection errors, but since they process all persons in a scene at once, they are prone to errors, particularly for persons in small scales. To address all these challenges, we propose the integration of top-down and bottom-up approaches to exploit their strengths. Our top-down network estimates human joints from all persons instead of one in an image patch, making it robust to possible erroneous bounding boxes. Our bottom-up network incorporates human-detection based normalized heatmaps, allowing the network to be more robust in handling scale variations. Finally, the estimated 3D poses from the top-down and bottom-up networks are fed into our integration network for final 3D poses. To address the common gaps between training and testing data, we do optimization during the test time, by refining the estimated 3D human poses using high-order temporal constraint, re-projection loss, and bone length regularizations. We also introduce a two-person pose discriminator that enforces natural two-person interactions. Finally, we apply a semi-supervised method to overcome the 3D ground-truth data scarcity. Our evaluations demonstrate the effectiveness of the proposed method and its individual components. Our code and pretrained models are available publicly: https://github.com/3dpose/3D-Multi-Person-Pose .
- Open Access
- 3 April 2020
- Published by Association for the Advancement of Artificial Intelligence (AAAI) in Proceedings of the AAAI Conference on Artificial Intelligence
- Vol. 34 (07) , 10631-10638
3D Human Pose Estimation Using Spatio-Temporal Networks with Explicit Occlusion Training
Estimating 3D poses from a monocular video is still a challenging task, despite the significant progress that has been made in the recent years. Generally, the performance of existing methods drops when the target person is too small/large, or the motion is too fast/slow relative to the scale and speed of the training data. Moreover, to our knowledge, many of these methods are not designed or trained under severe occlusion explicitly, making their performance on handling occlusion compromised. Addressing these problems, we introduce a spatio-temporal network for robust 3D human pose estimation. As humans in videos may appear in different scales and have various motion speeds, we apply multi-scale spatial features for 2D joints or keypoints prediction in each individual frame, and multi-stride temporal convolutional networks (TCNs) to estimate 3D joints or keypoints. Furthermore, we design a spatio-temporal discriminator based on body structures as well as limb motions to assess whether the predicted pose forms a valid pose and a valid movement. During training, we explicitly mask out some keypoints to simulate various occlusion cases, from minor to severe occlusion, so that our network can learn better and becomes robust to various degrees of occlusion. As there are limited 3D ground truth data, we further utilize 2D video data to inject a semi-supervised learning capability to our network. Experiments on public data sets validate the effectiveness of our method, and our ablation studies show the strengths of our network's individual submodules.
- 3 April 2020
- Published by Association for the Advancement of Artificial Intelligence (AAAI) in Proceedings of the AAAI Conference on Artificial Intelligence
- Vol. 34 (07) , 10631-10638
Open Access3D Human Pose Estimation Using Spatio-Temporal Networks with Explicit Occlusion Training
Estimating 3D poses from a monocular video is still a challenging task, despite the significant progress that has been made in the recent years. Generally, the performance of existing methods drops when the target person is too small/large, or the motion is too fast/slow relative to the scale and speed of the training data. Moreover, to our knowledge, many of these methods are not designed or trained under severe occlusion explicitly, making their performance on handling occlusion compromised. Addressing these problems, we introduce a spatio-temporal network for robust 3D human pose estimation. As humans in videos may appear in different scales and have various motion speeds, we apply multi-scale spatial features for 2D joints or keypoints prediction in each individual frame, and multi-stride temporal convolutional networks (TCNs) to estimate 3D joints or keypoints. Furthermore, we design a spatio-temporal discriminator based on body structures as well as limb motions to assess whether the predicted pose forms a valid pose and a valid movement. During training, we explicitly mask out some keypoints to simulate various occlusion cases, from minor to severe occlusion, so that our network can learn better and becomes robust to various degrees of occlusion. As there are limited 3D ground truth data, we further utilize 2D video data to inject a semi-supervised learning capability to our network. Experiments on public data sets validate the effectiveness of our method, and our ablation studies show the strengths of our network's individual submodules.
- 1 October 2019
- Published by Institute of Electrical and Electronics Engineers (IEEE)
- p. 723-732
Occlusion-Aware Networks for 3D Human Pose Estimation in Video
- 1 October 2019
- Published by Institute of Electrical and Electronics Engineers (IEEE)
- p. 723-732
Occlusion-Aware Networks for 3D Human Pose Estimation in Video
- 12 September 2017
- Published by Wiley in Medical Physics
- Vol. 44 (11) , 5889-5900
A block matching based approach with multiple simultaneous templates for the real‐time 2D ultrasound tracking of liver vessels
Purpose: The implementation of motion management techniques in radiation therapy can aid in mitigating uncertainties and reducing margins. For motion management to be effective, it is necessary to track key structures both accurately and at a real‐time speed. Therefore, the focus of this work was to develop a 2D algorithm for the real‐time tracking of ultrasound features to aid in radiation therapy motion management.Materials and Methods: The developed algorithm utilized a similarity measure‐based block matching algorithm incorporating training methods and multiple simultaneous templates. The algorithm is broken down into three primary components, all of which use normalized cross‐correlation (NCC) as a similarity metric. First, a global feature shift to account for gross displacements from the previous frame is determined using large block sizes which encompass the entirety of the feature. Second, the most similar reference frame is chosen from a series of training images that are accumulated during the first K frames of tracking to aid in contour consistency and provide a starting point for the localized template initialization. Finally, localized block matching is performed through the simultaneous use of both a training frame and the previous frame. The localized block matching utilizes a series of templates positioned at the boundary points of the training and previous contours. The weighted final boundary points from both the previous and the training frame are ultimately combined and used to determine an affine transformation from the previous frame to the current frame.Results: A mean tracking error of 0.72 ± 1.25 mm was observed for 85 point‐landmarks across 39 ultrasound sequences relative to manual ground truth annotations. The image processing speed per landmark with the GPU implementation was between 41 and 165 frames per second (fps) during the training set accumulation, and between 73 and 234 fps after training set accumulation. Relative to a comparable multithreaded CPU approach using OpenMP, the GPU implementation resulted in speedups between −30% and 355% during training set accumulation, and between −37% and 639% postaccumulation.Conclusions: Initial implementations indicated an accuracy that was comparable to or exceeding those achieved by alternative 2D tracking methods, with a computational speed that is more than sufficient for real‐time applications in a radiation therapy environment. While the overall performance reached levels suitable for implementation in radiation therapy, the observed increase in failures for smaller features, as well as the algorithm's inability to be applied to nonconvex features warrants additional investigation to address the shortcomings observed.
- 12 September 2017
- Published by Wiley in Medical Physics
- Vol. 44 (11) , 5889-5900
A block matching based approach with multiple simultaneous templates for the real‐time 2D ultrasound tracking of liver vessels
Purpose: The implementation of motion management techniques in radiation therapy can aid in mitigating uncertainties and reducing margins. For motion management to be effective, it is necessary to track key structures both accurately and at a real‐time speed. Therefore, the focus of this work was to develop a 2D algorithm for the real‐time tracking of ultrasound features to aid in radiation therapy motion management.Materials and Methods: The developed algorithm utilized a similarity measure‐based block matching algorithm incorporating training methods and multiple simultaneous templates. The algorithm is broken down into three primary components, all of which use normalized cross‐correlation (NCC) as a similarity metric. First, a global feature shift to account for gross displacements from the previous frame is determined using large block sizes which encompass the entirety of the feature. Second, the most similar reference frame is chosen from a series of training images that are accumulated during the first K frames of tracking to aid in contour consistency and provide a starting point for the localized template initialization. Finally, localized block matching is performed through the simultaneous use of both a training frame and the previous frame. The localized block matching utilizes a series of templates positioned at the boundary points of the training and previous contours. The weighted final boundary points from both the previous and the training frame are ultimately combined and used to determine an affine transformation from the previous frame to the current frame.Results: A mean tracking error of 0.72 ± 1.25 mm was observed for 85 point‐landmarks across 39 ultrasound sequences relative to manual ground truth annotations. The image processing speed per landmark with the GPU implementation was between 41 and 165 frames per second (fps) during the training set accumulation, and between 73 and 234 fps after training set accumulation. Relative to a comparable multithreaded CPU approach using OpenMP, the GPU implementation resulted in speedups between −30% and 355% during training set accumulation, and between −37% and 639% postaccumulation.Conclusions: Initial implementations indicated an accuracy that was comparable to or exceeding those achieved by alternative 2D tracking methods, with a computational speed that is more than sufficient for real‐time applications in a radiation therapy environment. While the overall performance reached levels suitable for implementation in radiation therapy, the observed increase in failures for smaller features, as well as the algorithm's inability to be applied to nonconvex features warrants additional investigation to address the shortcomings observed.
- 4 September 2017
- Published by Springer Nature
- p. 754-762
Data-Driven Rank Aggregation with Application to Grand Challenges
- 4 September 2017
- Published by Springer Nature
- p. 754-762
Data-Driven Rank Aggregation with Application to Grand Challenges
- 1 October 2016
- Published by Elsevier in Computer Vision and Image Understanding
- Vol. 151, 3-13
Modeling 4D pathological changes by leveraging normative models
- 1 October 2016
- Published by Elsevier in Computer Vision and Image Understanding
- Vol. 151, 3-13
Modeling 4D pathological changes by leveraging normative models
- 7 June 2016
- Published by Wiley in Medical Physics
- Vol. 43 (6) , 3635
SU-G-BRA-01: A Real-Time Tumor Localization and Guidance Platform for Radiotherapy Using US and MRI
- 7 June 2016
- Published by Wiley in Medical Physics
- Vol. 43 (6) , 3635
SU-G-BRA-01: A Real-Time Tumor Localization and Guidance Platform for Radiotherapy Using US and MRI