Shuyang Li
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
Shuyang Li has a total of 24 co-authors. Here are the top co-authors by number of shared publications:
- Wenqiang Chen
Department of Architecture and Civil Engineering, City University of Hong Kong, Hong Kong, China
8 shared publications Danny H W Li
5 shared publicationsDanny H.W. Li
Department of Architecture and Civil Engineering, Building Energy Research Group, City University of Hong Kong, Kowloon, Hong Kong
4 shared publicationsSiwei Lou
Department of Civil Engineering, Guangzhou University, Guangzhou, P.R. China
3 shared publicationsWenqiang Chen
Department of Architecture and Civil Engineering, Building Energy Research Group, City University of Hong Kong, Kowloon, Hong Kong
3 shared publications- Ernest Kin Wai Tsang
Hong Kong Metropolitan University
2 shared publications
Latest Publications
- Open Access
From Pixels to Predicates Structuring urban perception with scene graphs
Perception research is increasingly modelled using streetscapes, yet many approaches still rely on pixel features or object co-occurrence statistics, overlooking the explicit relations that shape human perception. This study proposes a three stage pipeline that transforms street view imagery (SVI) into structured representations for predicting six perceptual indicators. In the first stage, each image is parsed using an open-set Panoptic Scene Graph model (OpenPSG) to extract object predicate object triplets. In the second stage, compact scene-level embeddings are learned through a heterogeneous graph autoencoder (GraphMAE). In the third stage, a neural network predicts perception scores from these embeddings. We evaluate the proposed approach against image-only baselines in terms of accuracy, precision, and cross-city generalization. Results indicate that (i) our approach improves perception prediction accuracy by an average of 26% over baseline models, and (ii) maintains strong generalization performance in cross-city prediction tasks. Additionally, the structured representation clarifies which relational patterns contribute to lower perception scores in urban scenes, such as graffiti on wall and car parked on sidewalk. Overall, this study demonstrates that graph-based structure provides expressive, generalizable, and interpretable signals for modelling urban perception, advancing human-centric and context-aware urban analytics.
Open AccessFrom Pixels to Predicates Structuring urban perception with scene graphs
Perception research is increasingly modelled using streetscapes, yet many approaches still rely on pixel features or object co-occurrence statistics, overlooking the explicit relations that shape human perception. This study proposes a three stage pipeline that transforms street view imagery (SVI) into structured representations for predicting six perceptual indicators. In the first stage, each image is parsed using an open-set Panoptic Scene Graph model (OpenPSG) to extract object predicate object triplets. In the second stage, compact scene-level embeddings are learned through a heterogeneous graph autoencoder (GraphMAE). In the third stage, a neural network predicts perception scores from these embeddings. We evaluate the proposed approach against image-only baselines in terms of accuracy, precision, and cross-city generalization. Results indicate that (i) our approach improves perception prediction accuracy by an average of 26% over baseline models, and (ii) maintains strong generalization performance in cross-city prediction tasks. Additionally, the structured representation clarifies which relational patterns contribute to lower perception scores in urban scenes, such as graffiti on wall and car parked on sidewalk. Overall, this study demonstrates that graph-based structure provides expressive, generalizable, and interpretable signals for modelling urban perception, advancing human-centric and context-aware urban analytics.
- Open Access
Determination of Climate-Based Daylight Metrics under 15 CIE (International Commission on Illumination) Standard Skies and Three Representative Skies
Daylighting serves as a crucial sustainable strategy in assessing the built environment. Climate-based daylight metrics (CBDMs) have been widely employed to evaluate the dynamic performance of daylight. However, conventional CBDM calculations heavily rely on time-consuming full-scale computer simulations. In addition, the CBDMs need representative annual daylight data that are essential for CBDM analysis, and can pose challenges in many locations. Even when suitable daylight data are available, they may not accurately reflect current trends and conditions. This study aims to determine the various CBDMs using both the 15 CIE Standard Skies and the three representative skies specific to Hong Kong. All data were simulated from a software named RADIANCE (version 5.3). The R2s of the CBDMs under both the 15 CIE Standard Skies and the three representative skies were more than 0.89, and the MBEs and RMSEs were not more than 5.4% and 9.1%, respectively, when the outdoor illuminance measured in 2004 was employed. The findings could be adopted for other locations where the required daylight parameters were not systematically recorded.
Open AccessDetermination of Climate-Based Daylight Metrics under 15 CIE (International Commission on Illumination) Standard Skies and Three Representative Skies
Daylighting serves as a crucial sustainable strategy in assessing the built environment. Climate-based daylight metrics (CBDMs) have been widely employed to evaluate the dynamic performance of daylight. However, conventional CBDM calculations heavily rely on time-consuming full-scale computer simulations. In addition, the CBDMs need representative annual daylight data that are essential for CBDM analysis, and can pose challenges in many locations. Even when suitable daylight data are available, they may not accurately reflect current trends and conditions. This study aims to determine the various CBDMs using both the 15 CIE Standard Skies and the three representative skies specific to Hong Kong. All data were simulated from a software named RADIANCE (version 5.3). The R2s of the CBDMs under both the 15 CIE Standard Skies and the three representative skies were more than 0.89, and the MBEs and RMSEs were not more than 5.4% and 9.1%, respectively, when the outdoor illuminance measured in 2004 was employed. The findings could be adopted for other locations where the required daylight parameters were not systematically recorded.
- Open Access
- 1 January 2023
- Published by Education and Research in Computer Aided Architectural Design in Europe (eCAADe)
- Vol. 2, 431-440
Indispensable Effects of Surrounding Avatars in VR-based Wayfinding Experiments
Virtual reality technologies facilitate the presentation of the indoor and outdoor environment, enabling researchers to conduct environmental behaviour experiments and evidence-based design research with great feasibility. However, most previous environmental behaviour research based on VR technologies did not consider the surrounding crowd, especially wayfinding-related research. Thus, from a human-centric perspective, the credibility of these studies will be greatly argued. Are the avatars indispensable in the virtual environment? To answer this question, we designed a comparative test including three scenarios: the virtual environment without avatars, data-driven avatars in the virtual environment, and agent-based avatars in the virtual environment. Taking the Satellite Terminal of Shanghai Pudong International Airport as a study case, we developed an online VR wayfinding experiment platform. 435 participants were invited to this experiment, their wayfinding performance will be recorded and then analysed and visualisation. The results demonstrate that the presence or absence of avatars significantly impacts the participants' decision-making time in virtual environments. Besides, the distribution and movement of avatars may affect the participants' wayfinding behaviour. This study highlights the importance of avatars in VR-based experiments and validates the feasibility of data-driven avatars replicating real-world crowds.
- 1 January 2023
- Published by Education and Research in Computer Aided Architectural Design in Europe (eCAADe)
- Vol. 2, 431-440
Open AccessIndispensable Effects of Surrounding Avatars in VR-based Wayfinding Experiments
Virtual reality technologies facilitate the presentation of the indoor and outdoor environment, enabling researchers to conduct environmental behaviour experiments and evidence-based design research with great feasibility. However, most previous environmental behaviour research based on VR technologies did not consider the surrounding crowd, especially wayfinding-related research. Thus, from a human-centric perspective, the credibility of these studies will be greatly argued. Are the avatars indispensable in the virtual environment? To answer this question, we designed a comparative test including three scenarios: the virtual environment without avatars, data-driven avatars in the virtual environment, and agent-based avatars in the virtual environment. Taking the Satellite Terminal of Shanghai Pudong International Airport as a study case, we developed an online VR wayfinding experiment platform. 435 participants were invited to this experiment, their wayfinding performance will be recorded and then analysed and visualisation. The results demonstrate that the presence or absence of avatars significantly impacts the participants' decision-making time in virtual environments. Besides, the distribution and movement of avatars may affect the participants' wayfinding behaviour. This study highlights the importance of avatars in VR-based experiments and validates the feasibility of data-driven avatars replicating real-world crowds.
- Open Access
- 5 December 2022
- Published by Elsevier in Journal of Materials Research and Technology
- Vol. 22, 1423-1435
Strength and chloride resistance of mortars blended with SCBA: the effect of calcination and particle sizing on its pozzolanic activity
- 5 December 2022
- Published by Elsevier in Journal of Materials Research and Technology
- Vol. 22, 1423-1435
Open AccessStrength and chloride resistance of mortars blended with SCBA: the effect of calcination and particle sizing on its pozzolanic activity
- 9 March 2022
- Published by SAGE Publications in Indoor and Built Environment
- Vol. 31 (6) , 1700-1714
Simple correlations between point daylight factor, average daylight factor and vertical daylight factor under all sky conditions and building design implications
The customary method for calculating daylight illuminance in a building is the daylight factor approach, which is assumed under the conventional overcast sky. However, such an approach is not flexible enough to predict diffuse illuminance in the presence of non-overcast skies. The daylight factor is invariant to building orientation and cannot take realistic and time-varying climatic conditions into account. Daylight in buildings is estimated using computer simulation techniques. However, full-scale computer simulations can be costly and time-consuming. Practitioners welcome simple calculation aids established via comprehensive analysis. Such easy tools would give building professionals and students basic and concise insight into the independency of different daylight parameters. Recently, daylight factor calculations have been extended to non-overcast skies. It means that the daylight factor approach can be a dynamic metric. This paper presents the calculation of the point daylight factor, the average daylight factor and the vertical daylight factor under all sky conditions, as well as building façade design implications. The performance of the three types of daylight factor for a typical room at various scattering angles is elaborated and evaluated; simple correlations between them are developed.
- 9 March 2022
- Published by SAGE Publications in Indoor and Built Environment
- Vol. 31 (6) , 1700-1714
Simple correlations between point daylight factor, average daylight factor and vertical daylight factor under all sky conditions and building design implications
The customary method for calculating daylight illuminance in a building is the daylight factor approach, which is assumed under the conventional overcast sky. However, such an approach is not flexible enough to predict diffuse illuminance in the presence of non-overcast skies. The daylight factor is invariant to building orientation and cannot take realistic and time-varying climatic conditions into account. Daylight in buildings is estimated using computer simulation techniques. However, full-scale computer simulations can be costly and time-consuming. Practitioners welcome simple calculation aids established via comprehensive analysis. Such easy tools would give building professionals and students basic and concise insight into the independency of different daylight parameters. Recently, daylight factor calculations have been extended to non-overcast skies. It means that the daylight factor approach can be a dynamic metric. This paper presents the calculation of the point daylight factor, the average daylight factor and the vertical daylight factor under all sky conditions, as well as building façade design implications. The performance of the three types of daylight factor for a typical room at various scattering angles is elaborated and evaluated; simple correlations between them are developed.
- Open Access
- 3 March 2022
- Published by Frontiers Media SA in Frontiers in Sustainable Cities
- Vol. 4
Predicting Daylight Illuminance for 15 CIE Standard Skies Using a Simple Software Tool
Estimation of internal daylight availability is the key step in evaluating various daylighting schemes. Interior daylight is often expressed as daylight factor under the traditional International Commission on Illumination (CIE) overcast sky. Nonetheless, such approach may be inflexible to estimate the kinetic situations as the solar locations vary under cloudless skies. Recently, the CIE adopted 15 standard skies covering the usual skies found in the world. The interior daylight illuminance is influenced by the sky conditions in terms of luminance levels and distributions with respect to the window-facing orientation at any given period. The daylight coefficient concept considering the variations of the sky luminance patterns can be employed to accurately compute the indoor daylight level. A full-scale computer simulation can be costly in modeling the building geometry with a long simulation time. Moreover, additional times and advanced skills are required for preparation and post-processing. Simple design programs would be convenient and helpful in preliminary design phase when various designs and schemes are being studied and analyzed. This paper adopts a simple computer program for computing the interior daylight under the 15 CIE standard skies. The capability of the suggested software was assessed against the simulated results using a computer program, namely, RADIANCE, and the field measurement readings. The running time based on the simple program was far less than that using RADIANCE. The findings showed that the internal daylight simulated by the proposed simple software were in good agreement with the simulated results by RADIANCE and measured readings. The peak daylight factor discrepancy was <2%, and the RMSE was 14.5% of the corresponding measured average values.
- 3 March 2022
- Published by Frontiers Media SA in Frontiers in Sustainable Cities
- Vol. 4
Open AccessPredicting Daylight Illuminance for 15 CIE Standard Skies Using a Simple Software Tool
Estimation of internal daylight availability is the key step in evaluating various daylighting schemes. Interior daylight is often expressed as daylight factor under the traditional International Commission on Illumination (CIE) overcast sky. Nonetheless, such approach may be inflexible to estimate the kinetic situations as the solar locations vary under cloudless skies. Recently, the CIE adopted 15 standard skies covering the usual skies found in the world. The interior daylight illuminance is influenced by the sky conditions in terms of luminance levels and distributions with respect to the window-facing orientation at any given period. The daylight coefficient concept considering the variations of the sky luminance patterns can be employed to accurately compute the indoor daylight level. A full-scale computer simulation can be costly in modeling the building geometry with a long simulation time. Moreover, additional times and advanced skills are required for preparation and post-processing. Simple design programs would be convenient and helpful in preliminary design phase when various designs and schemes are being studied and analyzed. This paper adopts a simple computer program for computing the interior daylight under the 15 CIE standard skies. The capability of the suggested software was assessed against the simulated results using a computer program, namely, RADIANCE, and the field measurement readings. The running time based on the simple program was far less than that using RADIANCE. The findings showed that the internal daylight simulated by the proposed simple software were in good agreement with the simulated results by RADIANCE and measured readings. The peak daylight factor discrepancy was <2%, and the RMSE was 14.5% of the corresponding measured average values.
- 2 March 2022
- Published by Elsevier in Journal of Building Engineering
- Vol. 51
Estimation of sky and externally reflected components under various obstructed CIE skies
- 2 March 2022
- Published by Elsevier in Journal of Building Engineering
- Vol. 51
Estimation of sky and externally reflected components under various obstructed CIE skies