Baining Zhao
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
Baining Zhao has a total of 46 co-authors. Here are the top co-authors by number of shared publications:
- Xinlei Chen
Tsinghua University
5 shared publications - Chen Gao
Tsinghua University
4 shared publications - Yong Li
Tsinghua University
4 shared publications - Fanhang Man
Shenzhen International Graduate School, Tsinghua University, Shenzhen, China
3 shared publications - Weichen Zhang
Shenzhen International Graduate School, Tsinghua University, Shenzhen, China and Pengcheng Laboratory, Shenzhen, China
3 shared publications - Wenhu Tang
School of Electric Power Engineering, South China University of Technology, Guangzhou 510641, China
3 shared publications
Latest Publications
- Open Access
- 9 November 2026
- Published by Association for Computing Machinery (ACM)
- p. 16697-16706
Aerial World Model for Long-horizon Visual Generation and Navigation in 3D Space
- 9 November 2026
- Published by Association for Computing Machinery (ACM)
- p. 16697-16706
Open AccessAerial World Model for Long-horizon Visual Generation and Navigation in 3D Space
- Open Access
- 8 August 2026
- Published by Association for Computing Machinery (ACM)
- p. 10326-10337
How Far Are Large Multimodal Models from Human-Level Spatial Action? A Benchmark for Goal-Oriented Embodied Navigation in Urban Airspace
- 8 August 2026
- Published by Association for Computing Machinery (ACM)
- p. 10326-10337
Open AccessHow Far Are Large Multimodal Models from Human-Level Spatial Action? A Benchmark for Goal-Oriented Embodied Navigation in Urban Airspace
- Top Cited PapersOpen Access
- 27 October 2025
- Published by Association for Computing Machinery (ACM)
- p. 11071-11080
Embodied-R: Collaborative Framework for Activating Embodied Spatial Reasoning in Foundation Models via Reinforcement Learning
- 27 October 2025
- Published by Association for Computing Machinery (ACM)
- p. 11071-11080
Top Cited PapersOpen AccessEmbodied-R: Collaborative Framework for Activating Embodied Spatial Reasoning in Foundation Models via Reinforcement Learning
- Open Access
- 27 October 2025
- Published by Association for Computing Machinery (ACM)
- p. 12519-12528
AirScape: An Aerial Generative World Model with Motion Controllability
- 27 October 2025
- Published by Association for Computing Machinery (ACM)
- p. 12519-12528
Open AccessAirScape: An Aerial Generative World Model with Motion Controllability
- Open Access
- 8 May 2024
- Published by Springer Nature in Humanities and Social Sciences Communications
- Vol. 11 (1) , 1-15
Estimating and modeling spontaneous mobility changes during the COVID-19 pandemic without stay-at-home orders
Comprehending the complex interplay among urban mobility, human behavior, and the COVID-19 pandemic could deliver vital perspectives to steer forthcoming public health endeavors. In late 2022, China lifted its "Zero-COVID" policy and rapidly abandoned nearly all interventions. It provides a unique opportunity to observe spontaneous mobility changes without government restriction throughout such a pandemic with high infection. Based on 148 million travel data from the public bus, subway, and taxi systems in Shenzhen, China, our analysis reveals discernible spatial discrepancies within mobility patterns. This phenomenon can be ascribed to the heterogeneous responses of mobility behavior tailored to specific purposes and travel modes in reaction to the pandemic. Considering both the physiological effects of virus infection and subjective willingness to travel, a dynamic model is proposed and capable of fitting fine-grained urban mobility. The analysis and model can interpret mobility data and underlying population behavior to inform policymakers when evaluating public health strategies against future large-scale infectious diseases.
- 8 May 2024
- Published by Springer Nature in Humanities and Social Sciences Communications
- Vol. 11 (1) , 1-15
Open AccessEstimating and modeling spontaneous mobility changes during the COVID-19 pandemic without stay-at-home orders
Comprehending the complex interplay among urban mobility, human behavior, and the COVID-19 pandemic could deliver vital perspectives to steer forthcoming public health endeavors. In late 2022, China lifted its "Zero-COVID" policy and rapidly abandoned nearly all interventions. It provides a unique opportunity to observe spontaneous mobility changes without government restriction throughout such a pandemic with high infection. Based on 148 million travel data from the public bus, subway, and taxi systems in Shenzhen, China, our analysis reveals discernible spatial discrepancies within mobility patterns. This phenomenon can be ascribed to the heterogeneous responses of mobility behavior tailored to specific purposes and travel modes in reaction to the pandemic. Considering both the physiological effects of virus infection and subjective willingness to travel, a dynamic model is proposed and capable of fitting fine-grained urban mobility. The analysis and model can interpret mobility data and underlying population behavior to inform policymakers when evaluating public health strategies against future large-scale infectious diseases.
- Open Access
- 11 September 2022
- Published by Association for Computing Machinery (ACM)
- p. 409-414
TRACT: Towards Large-Scale Crowdsensing With High-Efficiency Swarm Path Planning
Unmanned Aerial Vehicle (UAV) sensing swarms have grown in popularity because of their advantages of low-cost, high mobility, and high maneuverability in crowdsensing. The effectiveness of UAV swarm sensing is determined by the path of each UAV in the swarm. However, it is challenging to do path planning for a UAV swarm to satisfy various data requirements in large-scale crowdsensing tasks due to the area scale and long-term planning requirements. To achieve this, we propose TRACT, TowaRds lArge-sCale crowdsensing wiTh high-efficiency swarm path planning algorithm. We expand the sensing coverage problem for large-scale modeling. Furthermore, we propose a policy matrix searching technique with a simulated annealing algorithm to address the complexity of long-term planning. Our experiment shows that TRACT adapts to various data requirements, with the performance improved by 90% in terms of KL-divergence between the time-aggregated data value and the data requirements, while time cost for path planning was reduced by 43% compared with the previous state-of-the-art method approach.
- 11 September 2022
- Published by Association for Computing Machinery (ACM)
- p. 409-414
Open AccessTRACT: Towards Large-Scale Crowdsensing With High-Efficiency Swarm Path Planning
Unmanned Aerial Vehicle (UAV) sensing swarms have grown in popularity because of their advantages of low-cost, high mobility, and high maneuverability in crowdsensing. The effectiveness of UAV swarm sensing is determined by the path of each UAV in the swarm. However, it is challenging to do path planning for a UAV swarm to satisfy various data requirements in large-scale crowdsensing tasks due to the area scale and long-term planning requirements. To achieve this, we propose TRACT, TowaRds lArge-sCale crowdsensing wiTh high-efficiency swarm path planning algorithm. We expand the sensing coverage problem for large-scale modeling. Furthermore, we propose a policy matrix searching technique with a simulated annealing algorithm to address the complexity of long-term planning. Our experiment shows that TRACT adapts to various data requirements, with the performance improved by 90% in terms of KL-divergence between the time-aggregated data value and the data requirements, while time cost for path planning was reduced by 43% compared with the previous state-of-the-art method approach.
- 4 November 2021
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Instrumentation and Measurement
- Vol. 70 (00189456) , 1-10
Signal-to-Signal Translation for Fault Diagnosis of Bearings and Gears With Few Fault Samples
Deep learning (DL)-based models can achieve satisfactory performance in bearing and gear fault diagnosis relying on sufficient fault vibration signals. However, it is difficult to acquire adequate machine fault data in industrial applications. Generally, data collected in the healthy condition are massive, but rare fault samples are available. To address the diagnostic problem with few fault data, this article proposes a method, namely, multilabel cycle translating adversarial network (MCTAN). The architecture can translate healthy vibration data into auxiliary fault data. Signal-to-signal translation enables auxiliary data to be acquired based on few fault data and adequate healthy data. Then, the auxiliary data are utilized to improve the performance of DL-based fault classification approaches. Experimental results and comparative analysis indicate the superiority and robustness of the proposed method for bearing and gear fault diagnosis on widely used XJTU-SY, Case Western Reserve University (CWRU), Jiangnan University (JNU), and Southeast University (SEU) datasets. Even if the ratio of healthy data to fault data ranges from 50:1 to 300:1, the proposed MCTAN succeeds in raising the diagnosis accuracy by up to 25.87% and outperforms other comparative methods.
- 4 November 2021
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Instrumentation and Measurement
- Vol. 70 (00189456) , 1-10
Signal-to-Signal Translation for Fault Diagnosis of Bearings and Gears With Few Fault Samples
Deep learning (DL)-based models can achieve satisfactory performance in bearing and gear fault diagnosis relying on sufficient fault vibration signals. However, it is difficult to acquire adequate machine fault data in industrial applications. Generally, data collected in the healthy condition are massive, but rare fault samples are available. To address the diagnostic problem with few fault data, this article proposes a method, namely, multilabel cycle translating adversarial network (MCTAN). The architecture can translate healthy vibration data into auxiliary fault data. Signal-to-signal translation enables auxiliary data to be acquired based on few fault data and adequate healthy data. Then, the auxiliary data are utilized to improve the performance of DL-based fault classification approaches. Experimental results and comparative analysis indicate the superiority and robustness of the proposed method for bearing and gear fault diagnosis on widely used XJTU-SY, Case Western Reserve University (CWRU), Jiangnan University (JNU), and Southeast University (SEU) datasets. Even if the ratio of healthy data to fault data ranges from 50:1 to 300:1, the proposed MCTAN succeeds in raising the diagnosis accuracy by up to 25.87% and outperforms other comparative methods.
- Open Access
- 13 August 2021
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Access
- Vol. 9, 125540-125549
Detection and Location of Safety Protective Wear in Power Substation Operation Using Wear-Enhanced YOLOv3 Algorithm
Wearing personal safety protective equipment (PSPE) plays a key role in reducing electrical injuries to electrical workers. However, substation employees often ignore this regulation due to lack of safety awareness and discomfortable feeling of wearing PSPE. Therefore, it is necessary to develop a detection algorithm for PSPE and workers to build real-time video surveillance systems in power substations. In this paper, a wear-enhanced YOLOv3 method for real-time detection of PSPE and substation workers is proposed. The gamma correction is first applied as the preprocessing method to highlight the details of the operators and data augmentation is performed. Next, K-means++ algorithm replaces K-means in wear-enhanced YOLOv3 method to derive the most suitable prior bounding box size and lift the detection speed. Then, the proposed method can be quickly and effectively trained based on transfer learning. Finally, extensive experiments are carried out on a dataset of images about usage of safety helmets, insulating gloves and boots. Using the proposed method, the mean average precision is improved by over 2% and the frames per second is the highest compared with other typical object detection methods, which illustrates the effectiveness of the wear-enhanced YOLOv3 method for PSPE and workers detection.
- 13 August 2021
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Access
- Vol. 9, 125540-125549
Open AccessDetection and Location of Safety Protective Wear in Power Substation Operation Using Wear-Enhanced YOLOv3 Algorithm
Wearing personal safety protective equipment (PSPE) plays a key role in reducing electrical injuries to electrical workers. However, substation employees often ignore this regulation due to lack of safety awareness and discomfortable feeling of wearing PSPE. Therefore, it is necessary to develop a detection algorithm for PSPE and workers to build real-time video surveillance systems in power substations. In this paper, a wear-enhanced YOLOv3 method for real-time detection of PSPE and substation workers is proposed. The gamma correction is first applied as the preprocessing method to highlight the details of the operators and data augmentation is performed. Next, K-means++ algorithm replaces K-means in wear-enhanced YOLOv3 method to derive the most suitable prior bounding box size and lift the detection speed. Then, the proposed method can be quickly and effectively trained based on transfer learning. Finally, extensive experiments are carried out on a dataset of images about usage of safety helmets, insulating gloves and boots. Using the proposed method, the mean average precision is improved by over 2% and the frames per second is the highest compared with other typical object detection methods, which illustrates the effectiveness of the wear-enhanced YOLOv3 method for PSPE and workers detection.
- Open Access
- 24 July 2020
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Access
- Vol. 8, 136487-136497
Electrical Equipment Identification Method With Synthetic Data Using Edge-Oriented Generative Adversarial Network
The fourth industrial revolution – Industry 4.0 – puts emphasis on the application of intelligent technologies in the area of monitoring and identification of electrical equipment. High precision and non-contact qualities make the infrared thermography one of the most suitable technologies for intelligent inspection of high-voltage apparatus. Yet, due to imperfect data acquisition methods and difficulties in collecting data, electrical equipment images are limited in quantities and imbalanced in representing different types of devices. Additionally, it is not easy to extract representative features of infrared images due to their low-intensity contrast and uneven distribution. In this paper, a data-driven framework is proposed for the identification of electrical equipment based on infrared images. The main technique of this proposed system is a novel process of generating synthetic infrared images. For this purpose, an Edge-Oriented Generative Adversarial Network (EOGAN) is developed. The EOGAN is designed to create realistic infrared images that can be used as augmented data for developing data-driven identification methods. Extracted edge features of electrical equipment are utilized as prior information to guide the process of generating realistic infrared images. Finally, comparative experiments are carried out to show the effectiveness of the proposed EOGAN-based framework for equipment identification in the presence of limited and imbalanced image datasets.
- 24 July 2020
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Access
- Vol. 8, 136487-136497
Open AccessElectrical Equipment Identification Method With Synthetic Data Using Edge-Oriented Generative Adversarial Network
The fourth industrial revolution – Industry 4.0 – puts emphasis on the application of intelligent technologies in the area of monitoring and identification of electrical equipment. High precision and non-contact qualities make the infrared thermography one of the most suitable technologies for intelligent inspection of high-voltage apparatus. Yet, due to imperfect data acquisition methods and difficulties in collecting data, electrical equipment images are limited in quantities and imbalanced in representing different types of devices. Additionally, it is not easy to extract representative features of infrared images due to their low-intensity contrast and uneven distribution. In this paper, a data-driven framework is proposed for the identification of electrical equipment based on infrared images. The main technique of this proposed system is a novel process of generating synthetic infrared images. For this purpose, an Edge-Oriented Generative Adversarial Network (EOGAN) is developed. The EOGAN is designed to create realistic infrared images that can be used as augmented data for developing data-driven identification methods. Extracted edge features of electrical equipment are utilized as prior information to guide the process of generating realistic infrared images. Finally, comparative experiments are carried out to show the effectiveness of the proposed EOGAN-based framework for equipment identification in the presence of limited and imbalanced image datasets.