Jun Yin
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Jun Yin has a total of 28 co-authors. Here are the top co-authors by number of shared publications:
- Pengyu Zeng
Shenzhen International Graduate School, Tsinghua University, Shenzhen 518055, China
5 shared publications - Shuai Lu
Tsinghua University
4 shared publications - Jing Zhong
Shenzhen International Graduate School, Tsinghua University, Shenzhen 518055, China
3 shared publications - Peilin Li
Department of Architecture, National University of Singapore, Singapore, Singapore
3 shared publications - Miao Zhang
Tsinghua University, Shenzhen, China
2 shared publications - Tianjing Feng
Institute for Environmental Design and Engineering, The Bartlett, University College London, London WC1E 6BT, UK
2 shared publications
Latest Publications
- Open Access
- 8 August 2026
- Published by Association for Computing Machinery (ACM)
- p. 343-353
BethAgent: A Contract-Governed Multi-Agent Framework for Verified Cytology Super-Resolution with Safe Rollback
- 8 August 2026
- Published by Association for Computing Machinery (ACM)
- p. 343-353
Open AccessBethAgent: A Contract-Governed Multi-Agent Framework for Verified Cytology Super-Resolution with Safe Rollback
- Open Access
ArchiExplain: Multi-Level Evidence Chains for Precedent-Based Interpretability in Architectural Image Understanding
Deep neural networks have been widely applied in architectural analysis and design research, supporting tasks such as facade recognition, floor-plan analysis, and architectural visual classification. However, although existing models possess strong predictive capabilities, their decision-making processes remain characterized by a pronounced black-box nature, making it difficult to provide architects with understandable and traceable grounds for judgment. This limits their practical value in the architectural field, as designers require not only accurate outputs but also interpretable explanatory evidence regarding the basis of decision-making. This issue is particularly critical in architectural interpretation, where judgments are rarely made solely on the basis of isolated visual features, but are instead often formed through comparison and negotiation with precedents, spatial logic, and domain knowledge. To address this challenge, this paper proposes ArchiExplain, a multi-level interpretability framework for architectural image understanding, aiming to enable a deeper understanding of architectural images. The main contributions of this study are threefold: (1) We construct two architectural datasets for interpretability evaluation: a facade dataset composed of streetscape images from Harbin, China, and Greece, and a floor-plan dataset consisting of Real-plan drawings from real design cases and standardized generated R-plan drawings. Unlike existing datasets that primarily serve style recognition, semantic parsing, or image generation tasks, the datasets in this paper focus on evaluating the correspondence among model explanations, precedent associations, visual evidence, and predictive judgments. (2) Based on the above datasets, we propose the ArchiExplain framework. Unlike attribution methods such as Grad-CAM, Saliency Maps, and Integrated Gradients, which mainly reveal local discriminative regions, or influence-based methods that only trace the influence of training samples, this framework integrates training-sample influence tracing, Saliency Maps, and Integrated Gradients. It establishes a unified evidential chain among precedent samples, discriminative image regions, and final predictions, thereby transforming neural network decisions into an interpretable reasoning process with architectural significance. (3) Experimental results show that ArchiExplain performs stably on 100 randomly selected test samples, achieving an accuracy of 98.41% in the facade classification task and 98.34% in the floor-plan classification task. Further deletion/occlusion faithfulness analysis shows that the main attribution methods outperform the random baseline. Meanwhile, a questionnaire study involving 28 architects further verifies the consistency between model explanations and human architectural cognition. These findings indicate that ArchiExplain can enhance the transparency of architectural deep learning models and has practical application potential in architectural design analysis, model diagnosis, and precedent-based learning.
Open AccessArchiExplain: Multi-Level Evidence Chains for Precedent-Based Interpretability in Architectural Image Understanding
Deep neural networks have been widely applied in architectural analysis and design research, supporting tasks such as facade recognition, floor-plan analysis, and architectural visual classification. However, although existing models possess strong predictive capabilities, their decision-making processes remain characterized by a pronounced black-box nature, making it difficult to provide architects with understandable and traceable grounds for judgment. This limits their practical value in the architectural field, as designers require not only accurate outputs but also interpretable explanatory evidence regarding the basis of decision-making. This issue is particularly critical in architectural interpretation, where judgments are rarely made solely on the basis of isolated visual features, but are instead often formed through comparison and negotiation with precedents, spatial logic, and domain knowledge. To address this challenge, this paper proposes ArchiExplain, a multi-level interpretability framework for architectural image understanding, aiming to enable a deeper understanding of architectural images. The main contributions of this study are threefold: (1) We construct two architectural datasets for interpretability evaluation: a facade dataset composed of streetscape images from Harbin, China, and Greece, and a floor-plan dataset consisting of Real-plan drawings from real design cases and standardized generated R-plan drawings. Unlike existing datasets that primarily serve style recognition, semantic parsing, or image generation tasks, the datasets in this paper focus on evaluating the correspondence among model explanations, precedent associations, visual evidence, and predictive judgments. (2) Based on the above datasets, we propose the ArchiExplain framework. Unlike attribution methods such as Grad-CAM, Saliency Maps, and Integrated Gradients, which mainly reveal local discriminative regions, or influence-based methods that only trace the influence of training samples, this framework integrates training-sample influence tracing, Saliency Maps, and Integrated Gradients. It establishes a unified evidential chain among precedent samples, discriminative image regions, and final predictions, thereby transforming neural network decisions into an interpretable reasoning process with architectural significance. (3) Experimental results show that ArchiExplain performs stably on 100 randomly selected test samples, achieving an accuracy of 98.41% in the facade classification task and 98.34% in the floor-plan classification task. Further deletion/occlusion faithfulness analysis shows that the main attribution methods outperform the random baseline. Meanwhile, a questionnaire study involving 28 architects further verifies the consistency between model explanations and human architectural cognition. These findings indicate that ArchiExplain can enhance the transparency of architectural deep learning models and has practical application potential in architectural design analysis, model diagnosis, and precedent-based learning.
- Open Access
FP-MAE: A Self-Supervised Model for Floorplan Generation with Incomplete Inputs
Floor plans are a central representational component of architectural design, operating in close relation to sections, elevations, and three-dimensional reasoning to support the production and understanding of architectural space. In this context, we address the bounded computational task of completing incomplete floor plan representations as a form of early-stage design assistance, rather than treating the floor plan as an isolated architectural object. Within this workflow, being able to automatically complete a floor plan from an unfinished draft is highly valuable because it allows architects to generate preliminary schemes more quickly, streamline early discussions, and reduce the repetitive workload involved in revisions. To meet this need, we present FP-MAE, a self-supervised learning framework designed for floor plan completion. This study proposes three core contributions: (1) We developed FloorplanNet, a dedicated dataset that includes 8000 floorplans consisting of both schematic line drawings and color-coded plans, providing diverse yet consistent examples of residential layouts. (2) On top of this dataset, FP-MAE applies the Masked Autoencoder (MAE) strategy. By deliberately masking sections of a plan and using a lightweight Vision Transformer (ViT) to reconstruct the missing regions, the model learns to capture the global structural patterns of floor plans from limited local information. (3) We evaluated FP-MAE across multiple masking scenarios and compared its performance with state-of-the-art baselines. Beyond controlled experiments, we also tested the model on real sketches produced during the early stages of design projects, which demonstrated its robustness under practical conditions. The results show that FP-MAE can produce complete plans that are both accurate and functionally coherent, even when starting from highly incomplete inputs. FP-MAE is a practical and scalable solution for automated floor plan generation. It can be integrated into design software as a supportive tool to speed up concept development and option exploration, and it also points toward broader opportunities for applying AI in architectural automation. While the current framework operates on two-dimensional plan representations, future extensions may integrate multi-view information such as sections or three-dimensional models to better reflect the relational nature of architectural design representations.
Open AccessFP-MAE: A Self-Supervised Model for Floorplan Generation with Incomplete Inputs
Floor plans are a central representational component of architectural design, operating in close relation to sections, elevations, and three-dimensional reasoning to support the production and understanding of architectural space. In this context, we address the bounded computational task of completing incomplete floor plan representations as a form of early-stage design assistance, rather than treating the floor plan as an isolated architectural object. Within this workflow, being able to automatically complete a floor plan from an unfinished draft is highly valuable because it allows architects to generate preliminary schemes more quickly, streamline early discussions, and reduce the repetitive workload involved in revisions. To meet this need, we present FP-MAE, a self-supervised learning framework designed for floor plan completion. This study proposes three core contributions: (1) We developed FloorplanNet, a dedicated dataset that includes 8000 floorplans consisting of both schematic line drawings and color-coded plans, providing diverse yet consistent examples of residential layouts. (2) On top of this dataset, FP-MAE applies the Masked Autoencoder (MAE) strategy. By deliberately masking sections of a plan and using a lightweight Vision Transformer (ViT) to reconstruct the missing regions, the model learns to capture the global structural patterns of floor plans from limited local information. (3) We evaluated FP-MAE across multiple masking scenarios and compared its performance with state-of-the-art baselines. Beyond controlled experiments, we also tested the model on real sketches produced during the early stages of design projects, which demonstrated its robustness under practical conditions. The results show that FP-MAE can produce complete plans that are both accurate and functionally coherent, even when starting from highly incomplete inputs. FP-MAE is a practical and scalable solution for automated floor plan generation. It can be integrated into design software as a supportive tool to speed up concept development and option exploration, and it also points toward broader opportunities for applying AI in architectural automation. While the current framework operates on two-dimensional plan representations, future extensions may integrate multi-view information such as sections or three-dimensional models to better reflect the relational nature of architectural design representations.
- 9 January 2026
- Published by Springer Nature
- p. 536-551
Prompt-Guided Region-Adaptive Enhancement for Aesthetic Low-Light Imaging
- 9 January 2026
- Published by Springer Nature
- p. 536-551
Prompt-Guided Region-Adaptive Enhancement for Aesthetic Low-Light Imaging
- Open Access
- 27 October 2025
- Published by Association for Computing Machinery (ACM)
- p. 11298-11307
MRED-14: A Benchmark for Low-Energy Residential Floor Plan Generation with 14 Flexible Inputs
- 27 October 2025
- Published by Association for Computing Machinery (ACM)
- p. 11298-11307
Open AccessMRED-14: A Benchmark for Low-Energy Residential Floor Plan Generation with 14 Flexible Inputs
- Open Access
Comprehensive and Dedicated Metrics for Evaluating AI-Generated Residential Floor Plans
In response to the growing importance of AI-driven residential design and the lack of dedicated evaluation metrics, we propose the Residential Floor Plan Assessment (RFP-A), a comprehensive framework tailored to architectural evaluation. RFP-A consists of multiple metrics that assess key aspects of floor plans, including room count compliance, spatial connectivity, room locations, and geometric features. It incorporates both rule-based comparisons and graph-based analysis to ensure design requirements are met. A comparison of RFP-A and existing metrics was conducted both qualitatively and quantitatively, and it was revealed that RFP-A provides more robust, interpretable, and computationally efficient assessments of the accuracy and diversity of generated plans. We evaluated the performance of six existing floor plan generation models using RFP-A, showing that, surprisingly, only HouseDiffusion and FloorplanDiffusion achieved accuracies above 90%, while other models scored below or around 60%. We further conducted a quantitative comparison of diversity, revealing that FloorplanDiffusion, HouseDiffusion, and HouseGAN each demonstrated strengths in different aspects—graph structure, spatial location, and room geometry, respectively—while no model achieved consistently high diversity across all dimensions. In addition, existing metrics can not reflect the quality of generated designs well, and the diversity of the generated designs depends on both the model input and structure. Our study not only enhances the assessment of generated floor plans but also aids architects in utilizing numerous generated designs effectively.
Open AccessComprehensive and Dedicated Metrics for Evaluating AI-Generated Residential Floor Plans
In response to the growing importance of AI-driven residential design and the lack of dedicated evaluation metrics, we propose the Residential Floor Plan Assessment (RFP-A), a comprehensive framework tailored to architectural evaluation. RFP-A consists of multiple metrics that assess key aspects of floor plans, including room count compliance, spatial connectivity, room locations, and geometric features. It incorporates both rule-based comparisons and graph-based analysis to ensure design requirements are met. A comparison of RFP-A and existing metrics was conducted both qualitatively and quantitatively, and it was revealed that RFP-A provides more robust, interpretable, and computationally efficient assessments of the accuracy and diversity of generated plans. We evaluated the performance of six existing floor plan generation models using RFP-A, showing that, surprisingly, only HouseDiffusion and FloorplanDiffusion achieved accuracies above 90%, while other models scored below or around 60%. We further conducted a quantitative comparison of diversity, revealing that FloorplanDiffusion, HouseDiffusion, and HouseGAN each demonstrated strengths in different aspects—graph structure, spatial location, and room geometry, respectively—while no model achieved consistently high diversity across all dimensions. In addition, existing metrics can not reflect the quality of generated designs well, and the diversity of the generated designs depends on both the model input and structure. Our study not only enhances the assessment of generated floor plans but also aids architects in utilizing numerous generated designs effectively.
- Open Access
TSCnet: A Text-driven Semantic-level Controllable Framework for Customized Low-Light Image Enhancement
Deep learning-based image enhancement methods show significant advantages in reducing noise and improving visibility in low-light conditions. These methods are typically based on one-to-one mapping, where the model learns a direct transformation from low light to specific enhanced images. Therefore, these methods are inflexible as they do not allow highly personalized mapping, even though an individual's lighting preferences are inherently personalized. To overcome these limitations, we propose a new light enhancement task and a new framework that provides customized lighting control through prompt-driven, semantic-level, and quantitative brightness adjustments. The framework begins by leveraging a Large Language Model (LLM) to understand natural language prompts, enabling it to identify target objects for brightness adjustments. To localize these target objects, the Retinex-based Reasoning Segment (RRS) module generates precise target localization masks using reflection images. Subsequently, the Text-based Brightness Controllable (TBC) module adjusts brightness levels based on the generated illumination map. Finally, an Adaptive Contextual Compensation (ACC) module integrates multi-modal inputs and controls a conditional diffusion model to adjust the lighting, ensuring seamless and precise enhancements accurately. Experimental results on benchmark datasets demonstrate our framework's superior performance at increasing visibility, maintaining natural color balance, and amplifying fine details without creating artifacts. Furthermore, its robust generalization capabilities enable complex semantic-level lighting adjustments in diverse open-world environments through natural language interactions.
Open AccessTSCnet: A Text-driven Semantic-level Controllable Framework for Customized Low-Light Image Enhancement
Deep learning-based image enhancement methods show significant advantages in reducing noise and improving visibility in low-light conditions. These methods are typically based on one-to-one mapping, where the model learns a direct transformation from low light to specific enhanced images. Therefore, these methods are inflexible as they do not allow highly personalized mapping, even though an individual's lighting preferences are inherently personalized. To overcome these limitations, we propose a new light enhancement task and a new framework that provides customized lighting control through prompt-driven, semantic-level, and quantitative brightness adjustments. The framework begins by leveraging a Large Language Model (LLM) to understand natural language prompts, enabling it to identify target objects for brightness adjustments. To localize these target objects, the Retinex-based Reasoning Segment (RRS) module generates precise target localization masks using reflection images. Subsequently, the Text-based Brightness Controllable (TBC) module adjusts brightness levels based on the generated illumination map. Finally, an Adaptive Contextual Compensation (ACC) module integrates multi-modal inputs and controls a conditional diffusion model to adjust the lighting, ensuring seamless and precise enhancements accurately. Experimental results on benchmark datasets demonstrate our framework's superior performance at increasing visibility, maintaining natural color balance, and amplifying fine details without creating artifacts. Furthermore, its robust generalization capabilities enable complex semantic-level lighting adjustments in diverse open-world environments through natural language interactions.
- Open Access
- 1 January 2025
- Published by Association for Computational Linguistics (ACL)
- p. 6640-6662
FloorPlan-LLaMa: Aligning Architects’ Feedback and Domain Knowledge in Architectural Floor Plan Generation
Jun Yin, Pengyu Zeng, Haoyuan Sun, Yuqin Dai, Han Zheng, Miao Zhang, Yachao Zhang, Shuai Lu. Proceedings of the 63rd Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers). 2025.
- 1 January 2025
- Published by Association for Computational Linguistics (ACL)
- p. 6640-6662
Open AccessFloorPlan-LLaMa: Aligning Architects’ Feedback and Domain Knowledge in Architectural Floor Plan Generation
Jun Yin, Pengyu Zeng, Haoyuan Sun, Yuqin Dai, Han Zheng, Miao Zhang, Yachao Zhang, Shuai Lu. Proceedings of the 63rd Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers). 2025.