Haowen 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
Haowen Wang has a total of 40 co-authors. Here are the top co-authors by number of shared publications:
- Feifei Feng
Midea Group, Shanghai, China
2 shared publications - Jeroen van Ameijde
School of Architecture, The Chinese University of Hong Kong
2 shared publications - Jian Tang
Midea Group, Shanghai, China
2 shared publications - Xiuquan Qiao
State Key Laboratory of Networking and Switching Technology, Beijing University of Posts and Telecommunications, Beijing, China
2 shared publications - Zhengping Che
Beijing Innovation Center of Humanoid Robotics, Beijing, China
2 shared publications - Zhiyuan Xu
Midea Group, Shanghai, China
2 shared publications
Latest Publications
- 1 January 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Multimedia
- Vol. 28, 17272-17283
Adversarial Attack for RGB-Event Based Visual Object Tracking
- 1 January 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Multimedia
- Vol. 28, 17272-17283
Adversarial Attack for RGB-Event Based Visual Object Tracking
- Open Access
- 9 August 2024
- Published by Frontiers Media SA in Frontiers in Pharmacology
- Vol. 15, 1448381
Integrating network pharmacology, molecular docking and experimental verification to reveal the mechanism of artesunate in inhibiting choroidal melanoma
Background: Artesunate (ART), a natural compound derived from Artemisia annua, has shown promising clinical potentials in the treatment of various tumors, but the exact mechanism is unclear. Choroidal melanoma (CM) is a major malignant ocular tumor in adults, known for its significant malignancy and poor prognosis, with limited efficacy in current treatments. This study explored the anti-CM effects and mechanisms of ART using a combination of network pharmacology, molecular docking and experimental validation.Methods: Potential targets of ART were screened in PubChem, Swiss Target Prediction and Traditional Chinese Medicine Systems Pharmacology (TCMSP) Database Analysis Platform databases, while target genes related to CM prognosis were selected from Online Mendelian Inheritance in Man (OMIM), GeneCards and DisGeNET databases. The intersection of these two groups of datasets yielded the target genes of ART involved in CM. Protein-protein interaction (PPI) network analysis of the intersecting targets, as well as Gene Ontology (GO) and Kyoto Encyclopedia of Genes and Genomes (KEGG) analyses, were conducted to identify core targets and critical pathways. Molecular docking methods were performed to predict the binding interactions between ART and core targets. The effects of ART on CM were evaluated through CCK8, colony formation, transwell, as well as flow cytometry assays to detect apoptosis, cell cycle, reactive oxygen species (ROS). Western blot (WB) assays were conducted to investigate the impact of ART on key proteins and pathways associated with CM. Finally, in vivo assays were conducted to further validate the effects of ART on subcutaneous tumors in nude mice.Results: Research has shown that key pathways and core targets for ART in treating CM were identified through a network pharmacology approach. Molecular docking results verified the strong binding affinity between ART and these core targets. The analysis and predicted results indicated that ART primarily exerted its effects on CM through various tumor-related pathways like apoptosis. The assays in vitro confirmed that ART significantly inhibited the proliferation and migration of CM cells. This was achieved by promoting apoptosis through activation of the p53 signaling pathway, causing cell cycle arrest at the G0/G1 phase by inhibiting the PI3K/AKT/mTOR signaling pathway and increasing the intracellular level of ROS by activating the NRF2/HO-1 signaling pathway. Additionally, the assays in vivo further validated the significant proliferation-inhibitory effect of ART on CM.Conclusion: This study, making the initial exploration, illustrated through network pharmacology combined with molecular docking and in vitro/in vivo assays, confirmed that ART exerted potential anti-cancer effects on CM by promoting apoptosis, inducing cell cycle arrest and increasing intracellular levels of ROS. These findings suggested that ART held significant therapeutic potential for CM.
- 9 August 2024
- Published by Frontiers Media SA in Frontiers in Pharmacology
- Vol. 15, 1448381
Open AccessIntegrating network pharmacology, molecular docking and experimental verification to reveal the mechanism of artesunate in inhibiting choroidal melanoma
Background: Artesunate (ART), a natural compound derived from Artemisia annua, has shown promising clinical potentials in the treatment of various tumors, but the exact mechanism is unclear. Choroidal melanoma (CM) is a major malignant ocular tumor in adults, known for its significant malignancy and poor prognosis, with limited efficacy in current treatments. This study explored the anti-CM effects and mechanisms of ART using a combination of network pharmacology, molecular docking and experimental validation.Methods: Potential targets of ART were screened in PubChem, Swiss Target Prediction and Traditional Chinese Medicine Systems Pharmacology (TCMSP) Database Analysis Platform databases, while target genes related to CM prognosis were selected from Online Mendelian Inheritance in Man (OMIM), GeneCards and DisGeNET databases. The intersection of these two groups of datasets yielded the target genes of ART involved in CM. Protein-protein interaction (PPI) network analysis of the intersecting targets, as well as Gene Ontology (GO) and Kyoto Encyclopedia of Genes and Genomes (KEGG) analyses, were conducted to identify core targets and critical pathways. Molecular docking methods were performed to predict the binding interactions between ART and core targets. The effects of ART on CM were evaluated through CCK8, colony formation, transwell, as well as flow cytometry assays to detect apoptosis, cell cycle, reactive oxygen species (ROS). Western blot (WB) assays were conducted to investigate the impact of ART on key proteins and pathways associated with CM. Finally, in vivo assays were conducted to further validate the effects of ART on subcutaneous tumors in nude mice.Results: Research has shown that key pathways and core targets for ART in treating CM were identified through a network pharmacology approach. Molecular docking results verified the strong binding affinity between ART and these core targets. The analysis and predicted results indicated that ART primarily exerted its effects on CM through various tumor-related pathways like apoptosis. The assays in vitro confirmed that ART significantly inhibited the proliferation and migration of CM cells. This was achieved by promoting apoptosis through activation of the p53 signaling pathway, causing cell cycle arrest at the G0/G1 phase by inhibiting the PI3K/AKT/mTOR signaling pathway and increasing the intracellular level of ROS by activating the NRF2/HO-1 signaling pathway. Additionally, the assays in vivo further validated the significant proliferation-inhibitory effect of ART on CM.Conclusion: This study, making the initial exploration, illustrated through network pharmacology combined with molecular docking and in vitro/in vivo assays, confirmed that ART exerted potential anti-cancer effects on CM by promoting apoptosis, inducing cell cycle arrest and increasing intracellular levels of ROS. These findings suggested that ART held significant therapeutic potential for CM.
- 12 April 2024
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Pattern Analysis and Machine Intelligence
- Vol. 46 (11) , 7088-7101
RDFC-GAN: RGB-Depth Fusion CycleGAN for Indoor Depth Completion
Raw depth images captured in indoor scenarios frequently exhibit extensive missing values due to the inherent limitations of the sensors and environments. For example, transparent materials frequently elude detection by depth sensors; surfaces may introduce measurement inaccuracies due to their polished textures, extended distances, and oblique incidence angles from the sensor. The presence of incomplete depth maps imposes significant challenges for subsequent vision applications, prompting the development of numerous depth completion techniques to mitigate this problem. Numerous methods excel at reconstructing dense depth maps from sparse samples, but they often falter when faced with extensive contiguous regions of missing depth values, a prevalent and critical challenge in indoor environments. To overcome these challenges, we design a novel two-branch end-to-end fusion network named RDFC-GAN, which takes a pair of RGB and incomplete depth images as input to predict a dense and completed depth map. The first branch employs an encoder-decoder structure, by adhering to the Manhattan world assumption and utilizing normal maps from RGB-D information as guidance, to regress the local dense depth values from the raw depth map. The other branch applies an RGB-depth fusion CycleGAN, adept at translating RGB imagery into detailed, textured depth maps while ensuring high fidelity through cycle consistency. We fuse the two branches via adaptive fusion modules named W-AdaIN and train the model with the help of pseudo depth maps. Comprehensive evaluations on NYU-Depth V2 and SUN RGB-D datasets show that our method significantly enhances depth completion performance particularly in realistic indoor settings.
- 12 April 2024
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Pattern Analysis and Machine Intelligence
- Vol. 46 (11) , 7088-7101
RDFC-GAN: RGB-Depth Fusion CycleGAN for Indoor Depth Completion
Raw depth images captured in indoor scenarios frequently exhibit extensive missing values due to the inherent limitations of the sensors and environments. For example, transparent materials frequently elude detection by depth sensors; surfaces may introduce measurement inaccuracies due to their polished textures, extended distances, and oblique incidence angles from the sensor. The presence of incomplete depth maps imposes significant challenges for subsequent vision applications, prompting the development of numerous depth completion techniques to mitigate this problem. Numerous methods excel at reconstructing dense depth maps from sparse samples, but they often falter when faced with extensive contiguous regions of missing depth values, a prevalent and critical challenge in indoor environments. To overcome these challenges, we design a novel two-branch end-to-end fusion network named RDFC-GAN, which takes a pair of RGB and incomplete depth images as input to predict a dense and completed depth map. The first branch employs an encoder-decoder structure, by adhering to the Manhattan world assumption and utilizing normal maps from RGB-D information as guidance, to regress the local dense depth values from the raw depth map. The other branch applies an RGB-depth fusion CycleGAN, adept at translating RGB imagery into detailed, textured depth maps while ensuring high fidelity through cycle consistency. We fuse the two branches via adaptive fusion modules named W-AdaIN and train the model with the help of pseudo depth maps. Comprehensive evaluations on NYU-Depth V2 and SUN RGB-D datasets show that our method significantly enhances depth completion performance particularly in realistic indoor settings.
- 17 November 2023
- Published by Institute of Electrical and Electronics Engineers (IEEE)
- p. 5561-5564
Detection method for driver's gaze direction based on improved VGG network
Gaze detection is the basis for driver state monitoring and intent recognition as well as vision-based human-vehicle interaction. This paper proposed an appearance-based gaze direction vector detection model, taking the eye image, head posture and pupil center position as sample data and the gaze direction vector as sample labels, and using the improved VGG deep learning network for training to obtain the driver's gaze direction vector, and completing the gaze direction detection of the driver's gazing at different target regions. In the real car environment, we carried out experiments to recognize 6 different regions of the windshield, and the results show that the gaze direction detection method proposed in this paper is basically unaffected by changes in head posture and driving environment illumination, and the average detection accuracy for 10 drivers is 96.53%.
- 17 November 2023
- Published by Institute of Electrical and Electronics Engineers (IEEE)
- p. 5561-5564
Detection method for driver's gaze direction based on improved VGG network
Gaze detection is the basis for driver state monitoring and intent recognition as well as vision-based human-vehicle interaction. This paper proposed an appearance-based gaze direction vector detection model, taking the eye image, head posture and pupil center position as sample data and the gaze direction vector as sample labels, and using the improved VGG deep learning network for training to obtain the driver's gaze direction vector, and completing the gaze direction detection of the driver's gazing at different target regions. In the real car environment, we carried out experiments to recognize 6 different regions of the windshield, and the results show that the gaze direction detection method proposed in this paper is basically unaffected by changes in head posture and driving environment illumination, and the average detection accuracy for 10 drivers is 96.53%.
- Open Access
- 26 October 2023
- Published by Association for Computing Machinery (ACM)
- p. 3676-3685
DTF-Net: Category-Level Pose Estimation and Shape Reconstruction via Deformable Template Field
- 26 October 2023
- Published by Association for Computing Machinery (ACM)
- p. 3676-3685
Open AccessDTF-Net: Category-Level Pose Estimation and Shape Reconstruction via Deformable Template Field
- Open Access
- 1 January 2023
- Published by Education and Research in Computer Aided Architectural Design in Europe (eCAADe)
- Vol. 2, 451-460
Quantifying the Impact of Pedestrian-level Greening on the Perception of Walking
With the ongoing densification of cities, there is an increasing need to assess the impact of compressed urban spaces on residents' wellbeing. Sidewalks are a platform for connectivity and social interaction. However, city planners often focus on pragmatic aspects while designing sidewalk spaces and neglect pedestrians' walking experiences. Hong Kong is recognized as a high-density city with a high proportion of walking connections, but quality issues such as obstacles and pollution are systemic. Our study is centred around a pedestrian perception-oriented walking greening index and using Hong Kong as a case study site. Firstly, the Green View Index (GVI) was obtained through the results of semantic segmentation of Street View Imagery (SVI). The amount of planting near the sidewalk was calibrated by the Normalized Difference Vegetation Index (NDVI). Secondly, a questionnaire survey was conducted to obtain the preferences of pedestrians for the above two indicators. Its results were used to determine the weights of each indicator and obtain the final walking greening index. The research provides practical guidance for transforming sidewalk environments into multifunctional social spaces, providing community infrastructure in disadvantaged neighbourhoods. It contributes new urban design insights around the balance between urban construction and urban greening.
- 1 January 2023
- Published by Education and Research in Computer Aided Architectural Design in Europe (eCAADe)
- Vol. 2, 451-460
Open AccessQuantifying the Impact of Pedestrian-level Greening on the Perception of Walking
With the ongoing densification of cities, there is an increasing need to assess the impact of compressed urban spaces on residents' wellbeing. Sidewalks are a platform for connectivity and social interaction. However, city planners often focus on pragmatic aspects while designing sidewalk spaces and neglect pedestrians' walking experiences. Hong Kong is recognized as a high-density city with a high proportion of walking connections, but quality issues such as obstacles and pollution are systemic. Our study is centred around a pedestrian perception-oriented walking greening index and using Hong Kong as a case study site. Firstly, the Green View Index (GVI) was obtained through the results of semantic segmentation of Street View Imagery (SVI). The amount of planting near the sidewalk was calibrated by the Normalized Difference Vegetation Index (NDVI). Secondly, a questionnaire survey was conducted to obtain the preferences of pedestrians for the above two indicators. Its results were used to determine the weights of each indicator and obtain the final walking greening index. The research provides practical guidance for transforming sidewalk environments into multifunctional social spaces, providing community infrastructure in disadvantaged neighbourhoods. It contributes new urban design insights around the balance between urban construction and urban greening.
- Open Access
- 1 January 2023
- Published by Education and Research in Computer Aided Architectural Design in Europe (eCAADe)
- Vol. 2, 341-350
Data-Driven Urban Research Tools Within Design Education: New Perspectives on Complex Urban Data and Processes
- 1 January 2023
- Published by Education and Research in Computer Aided Architectural Design in Europe (eCAADe)
- Vol. 2, 341-350
Open AccessData-Driven Urban Research Tools Within Design Education: New Perspectives on Complex Urban Data and Processes
- 26 November 2021
- Published by Institute of Electrical and Electronics Engineers (IEEE)
- p. 281-286
Deep learning based pose estimation method using 3D point clouds
Pose estimation refers to the acquisition of a rigid transformation of an object relative to its original model coordinate system. This paper proposes a deep learning based approach for pose estimation with point clouds of textureless objects. The contribution of this paper can be summarized as follows: (1) A multi-scale local feature aggregation strategy for emphasizing the neighbor region of interest points. (2) The extension of the original spatial transformer network to point clouds, pose estimation and object classification as the output in a single proposed network. (3) A deep learning model combined with the symmetric function and the multi-scale features to improve the accuracy and robustness of the network model, and a new defined joint loss function by considering the objectives of pose estimation and classification. The experiments are conducted to verify the performance of pose estimation when the point clouds of textureless objects are taken as the input data, which shows that the proposed deep learning framework effectively performs on the pose estimation for both synthetic and real point clouds according to the experimental results.
- 26 November 2021
- Published by Institute of Electrical and Electronics Engineers (IEEE)
- p. 281-286
Deep learning based pose estimation method using 3D point clouds
Pose estimation refers to the acquisition of a rigid transformation of an object relative to its original model coordinate system. This paper proposes a deep learning based approach for pose estimation with point clouds of textureless objects. The contribution of this paper can be summarized as follows: (1) A multi-scale local feature aggregation strategy for emphasizing the neighbor region of interest points. (2) The extension of the original spatial transformer network to point clouds, pose estimation and object classification as the output in a single proposed network. (3) A deep learning model combined with the symmetric function and the multi-scale features to improve the accuracy and robustness of the network model, and a new defined joint loss function by considering the objectives of pose estimation and classification. The experiments are conducted to verify the performance of pose estimation when the point clouds of textureless objects are taken as the input data, which shows that the proposed deep learning framework effectively performs on the pose estimation for both synthetic and real point clouds according to the experimental results.
- 24 October 2016
- Published by Institute of Electrical and Electronics Engineers (IEEE)
- p. 2081-2085
Design of hardware-in-the-loop simulation system for LTE-A system based on USRP
The traditional LTE-A algorithm validation is limited to pure computer software simulation, which can't simulate the actual link scenes, missing scene parameters, the simulation results is invaluable in practice. Therefore, this paper designs LTE-A hardware-in-the-loop simulation system based on software-defined radio device USRP, and the physical layer real-time test level downlink is designed on the system. The downlink scheduled by the scheduling algorithm of the PC simulation platform, with scene parameters feedback to MAC layer in real time. The downlink designed with power interference between cells and users simulate the actual scene. Compare to current design of LTE hardware-in-the-loop simulation, this design presents data interaction between MAC layer and physical layer, implement of specific requirements for physical layer link, instead of generation of vector signal only. The results show that the system is running normally with easy scalability.
- 24 October 2016
- Published by Institute of Electrical and Electronics Engineers (IEEE)
- p. 2081-2085
Design of hardware-in-the-loop simulation system for LTE-A system based on USRP
The traditional LTE-A algorithm validation is limited to pure computer software simulation, which can't simulate the actual link scenes, missing scene parameters, the simulation results is invaluable in practice. Therefore, this paper designs LTE-A hardware-in-the-loop simulation system based on software-defined radio device USRP, and the physical layer real-time test level downlink is designed on the system. The downlink scheduled by the scheduling algorithm of the PC simulation platform, with scene parameters feedback to MAC layer in real time. The downlink designed with power interference between cells and users simulate the actual scene. Compare to current design of LTE hardware-in-the-loop simulation, this design presents data interaction between MAC layer and physical layer, implement of specific requirements for physical layer link, instead of generation of vector signal only. The results show that the system is running normally with easy scalability.