Yu Liu’s research while affiliated with Tsinghua University and other places

What is this page?


This page lists works of an author who doesn't have a ResearchGate profile or hasn't added the works to their profile yet. It is automatically generated from public (personal) data to further our legitimate goal of comprehensive and accurate scientific recordkeeping. If you are this author and want this page removed, please let us know.

Publications (833)


YOLOv7scb: A Small-Target Object Detection Method for Fire Smoke Inspection
  • Article
  • Full-text available

February 2025

·

3 Reads

Fire

Dan Shao

·

Yu Liu

·

Guoxing Liu

·

[...]

·

Guangmin Liang

Fire detection presents considerable challenges due to the destructive and unpredictable characteristics of fires. These difficulties are amplified by the small size and low-resolution nature of fire and smoke targets in images captured from a distance, making it hard for models to extract relevant features. To address this, we introduce a novel method for small-target fire and smoke detection named YOLOv7scb. This approach incorporates two key improvements to the YOLOv7 framework: the use of space-to-depth convolution (SPD-Conv) and C3 modules, enhancing the model’s ability to extract features from small targets effectively. Additionally, a weighted bidirectional feature pyramid network (BiFPN) is integrated into the feature-extraction network to merge features across scales efficiently without increasing the model’s complexity. We also replace the conventional complete intersection over union (CIoU) loss function with Focal-CIoU, which reduces the degrees of freedom in the loss function and improves the model’s robustness. Given the limited size of the initial fire and smoke dataset, a transfer-learning strategy is applied during training. Experimental results demonstrate that our proposed model surpasses others in metrics such as precision and recall. Notably, it achieves a precision of 98.8% for small-target flame detection and 90.6% for small-target smoke detection. These findings underscore the model’s effectiveness and its broad potential for fire detection and mitigation applications.

Download



MetaCity: Data-driven sustainable development of complex cities

February 2025

·

46 Reads

The Innovation

Cities are complex systems that develop under complicated interactions among their human and environmental components. Urbanization generates substantial outcomes and opportunities while raising challenges including congestion, air pollution, inequality, etc., calling for efficient and reasonable solutions to sustainable developments. Fortunately, booming technologies generate large-scale data of complex cities, providing a chance to propose data-driven solutions for sustainable urban developments. This paper provides a comprehensive overview of data-driven urban sustainability practice. In this review article, we conceptualize MetaCity, a general framework for optimizing resource usage and allocation problems in complex cities with data-driven approaches. Under this framework, we decompose specific urban sustainable goals, e.g., efficiency and resilience, review practical urban problems under these goals, and explore the probability of using data-driven technologies as potential solutions to the challenge of complexity. On the basis of extensive urban data, we integrate urban problem discovery, operation of urban systems simulation, and complex decision-making problem solving into an entire cohesive framework to achieve sustainable development goals by optimizing resource allocation problems in complex cities.




RegionGCN: Spatial-Heterogeneity-Aware Graph Convolutional Networks

January 2025

·

28 Reads

Modeling spatial heterogeneity in the data generation process is essential for understanding and predicting geographical phenomena. Despite their prevalence in geospatial tasks, neural network models usually assume spatial stationarity, which could limit their performance in the presence of spatial process heterogeneity. By allowing model parameters to vary over space, several approaches have been proposed to incorporate spatial heterogeneity into neural networks. However, current geographically weighting approaches are ineffective on graph neural networks, yielding no significant improvement in prediction accuracy. We assume the crux lies in the over-fitting risk brought by a large number of local parameters. Accordingly, we propose to model spatial process heterogeneity at the regional level rather than at the individual level, which largely reduces the number of spatially varying parameters. We further develop a heuristic optimization procedure to learn the region partition adaptively in the process of model training. Our proposed spatial-heterogeneity-aware graph convolutional network, named RegionGCN, is applied to the spatial prediction of county-level vote share in the 2016 US presidential election based on socioeconomic attributes. Results show that RegionGCN achieves significant improvement over the basic and geographically weighted GCNs. We also offer an exploratory analysis tool for the spatial variation of non-linear relationships through ensemble learning of regional partitions from RegionGCN. Our work contributes to the practice of Geospatial Artificial Intelligence (GeoAI) in tackling spatial heterogeneity.


Emergence of the Traffic Autonomous Zone (TAZ) for Telecommunication Operations from Spatial Heterogeneity in Cellular Networks

January 2025

·

2 Reads

In the field of telecommunications, various operations are driven by different physical quantities. Each has its own patterns in time and space, but all show some clustered structures in their spatial distribution. This reflects a unified rule of human mobility, suggesting the consistency among different telecommunication regionalization objectives. With this in mind, regionalization can be used to identify these patterns and can be applied to improve management efficiency in the context of "autonomous networks". This article introduces the "Traffic Autonomous Zone (TAZ)" concept. This approach aims to create a reasonable unified regionalization scheme by identifying spatial clusters. It is not just a practical way to partition cities based on telecommunications needs, but it also captures self-organization structure of cities in essence. We present examples of this regionalization method using real data. Compared to the popular Louvain community detection method, our approach is on the Pareto frontier, allowing for a balance among various metrics in telecommunications.


Urban sensing in the era of large language models

January 2025

·

427 Reads

·

1 Citation

The Innovation

Urban sensing has become increasingly important as cities evolve into the centers of human activities. Large language models (LLMs) offer new opportunities for urban sensing based on commonsense and worldview that emerged through their language-centric framework. This paper illustrates the transformative impact of LLMs, particularly in the potential of advancing next-generation urban sensing for exploring urban mechanisms. The discussion navigates through several key aspects, including enhancing knowledge transfer between humans and LLM, urban mechanisms awareness, and achieve automated decision-making with LLM agents. We emphasize the potential of LLMs to revolutionize urban sensing, offering a more comprehensive, efficient, and in-depth understanding of urban dynamics, and also acknowledge challenges in multi-modal data utilization, spatial-temporal cognition, cultural adaptability, and privacy preservation. The future of urban sensing with LLMs lies in leveraging their emerged intelligent and addressing these challenges to achieve more intelligent, responsible, and sustainable urban development.


Urban food delivery services as extreme heat adaptation

January 2025

·

83 Reads

Nature Cities

More frequent global extreme heat events prompt behavioral adaptations, such as reducing outdoor activities to relieve potential distress. The emergence of innovative daily life services in cities offers new avenues for implementing such adaptive strategies. Here we investigate whether urban residents augment food delivery consumption as an immediate response to hot weather in China. Analyzing extensive food delivery service data across 100 Chinese cities from 2017 to 2023, we observe a significant surge in lunchtime orders, exceeding 12.6%, as temperatures escalate from 20 °C to 35 °C, and reaching 21.4% at 40 °C. These increments, coupled with reduced heat exposure via food delivery services, are more pronounced among female, high-income and older individuals, signifying varying degrees of benefit among consumers. We further reveal the transfer of heat exposure from consumers to delivery riders, highlighting the gains and pains introduced by food delivery services and the need for policy intervention. The quantification and findings in this study provide unique insights for the design of efficient policies promoting extreme heat adaptation and ensuring social equity while fighting climate change.


Citations (14)


... Recently, the rapid advancement of deep learning in the domain of computer vision has significantly transformed the process of automated understanding of realistic images [36][37][38][39][40], and it has found widespread application across diverse professional domains [41][42][43][44][45]. Nevertheless, research into adversarial attacks on realistic images for deep learning is also progressing [46,47]. ...

Reference:

A Local Adversarial Attack with a Maximum Aggregated Region Sparseness Strategy for 3D Objects
AllSpark: A Multimodal Spatio-Temporal General Intelligence Model with Ten Modalities via Language as a Reference Framework
  • Citing Article
  • January 2025

IEEE Transactions on Geoscience and Remote Sensing

... Obstruction is fundamentally a problem of the relative position between the camera and the target element instances. While SVI can be collected using relatively standardized procedures, the distance and angle between environmental elements and the camera lens can vary significantly (Zou and Wang, 2022;Lumnitz et al., 2021;Huang et al., 2025). In studies focused on building assessments, some retrieved SVI may only show the sides or partial views of buildings due to the variation of horizontal angles, which might not provide sufficient façade features for classification or evaluation tasks (Zou and Wang, 2022). ...

No “true” greenery: Deciphering the bias of satellite and street view imagery in urban greenery measurement
  • Citing Article
  • December 2024

Building and Environment

... Within metro transfer contexts, a station's accessibility and attractiveness can influence usage at nearby stations. Without considering spatial dependence, models may miss neighboring effects and regional interactions [11]. We find that incorporating spatial dependence to explore nonlinear relationships between the built environment and bike-metro transfer facilitates a more accurate understanding of urban commuting complexity [12]. ...

Integrating smart card records and dockless bike-sharing data to understand the effect of the built environment on cycling as a feeder mode for metro trips
  • Citing Article
  • December 2024

Journal of Transport Geography

... Investigations have spotlighted the fact that the extracellular matrix (ECM) serves as a vital conduit for communication, potentially influencing behavior stress regulation and depression [29,30], with the intertwined connection between the ECM and immune processes [31,32]. The ECM has also been definitively shown to hold a crucial role in orchestrating inflammatory and neuropathic pain, as corroborated by multiple studies [33][34][35]. ...

Mechanistic study of celastrol-mediated inhibition of proinflammatory activation of macrophages in IgA nephropathy via down-regulating ECM1

International Journal of Biological Sciences

... The protective mechanism of Picein involves activating the Nrf2/Heme oxygenase 1 (HO-1)/GPX4 pathway. HO-1, regulated by Nrf2 and associated with antioxidant proteins, helps counter oxidative stress by facilitating Nrf2 nuclear translocation to bind the antioxidant response element of the HMOX1 gene [127][128][129][130][131]. Activation of the Nrf2/HO-1 pathway increases GPX4 levels to counter OS [132][133], highlighting the importance of the Nrf2/HO-1/GPX4 axis in countering BMSC ferroptosis. This class of antioxidants similarly exerts ferroptosis-inhibiting effects by activating intracellular antioxidant pathways. ...

Iridium(III) Photosensitizers Induce Simultaneous Pyroptosis and Ferroptosis for Multi‐Network Synergistic Tumor Immunotherapy
  • Citing Article
  • October 2024

Angewandte Chemie

... Even with the significant development in heat pipes, their theoretical modeling has a considerable complexity, mainly arising from the processes' convective and phase change characteristics, often leading to limitations in the feasibility of using these devices [20][21][22]. Mathematical/numerical models to predict the temperature distribution of heat pipes can present a 4% [23] to 30 K [24,25] result, which is different from the experimental results. For CFD simulations, the wall temperature can vary from 11.1 to 37.7 K [26], and the heat flux can have an average relative difference of up to 59.9% [27]. ...

Experimental study of heat pipe start-up characteristics and development of an enhanced model considering gas diffusion effects
  • Citing Article
  • September 2024

Applied Thermal Engineering

... Zhou et al. [26] integrate a Transformer block into the backbone network to enhance contextual information, capture long-range dependencies, and strengthen the network's competence to locate multi-scale vessels. Qin et al. [27] and Li et al. [28] developed an end-to-end and a globalto-local detection transformer network to enhance the performance of small ship target identification in SAR images. Feng et al. [29] utilized a lightweight Vision Transfomer (ViT), and Yang et al. [30] employed a Swin Transformer for multi-scale ship target identification from complex SAR images. ...

A Novel End-To-End Transformer Network for Small Scale Ship Detection in SAR Images
  • Citing Conference Paper
  • July 2024

... A CNN model which learns from both text and graph is suggested in [11] to leverage interactions with sequence features. Several Deep Neural Networks (DNN) models are developed to detect the regulatory elements [12], DNA enhancers [13] , regulatory variants in brain cells [14] and histone marks and predict gene expression [15]. ...

CatLearning: highly accurate gene expression prediction from histone mark

Briefings in Bioinformatics

... Recent advances in biological therapies offer potential solutions in this area, providing new hope for more effective interventions. Techniques or materials such as stem cell and exosome therapies, growth factor injections, annulus fibrosus repair, tissue engineering, biocompatible interbody cages made from polyetheretherketone (PEEK), 3D-printed implants and nanoparticle drug delivery systems were designed to promote the regeneration and repair of intervertebral discs (122)(123)(124)(125)(126)(127). Yu et al (122) demonstrate that transplanting of menstrual blood-derived stem cells embedded in collagen I gel into annulus fibrosus defects after discectomy in rats preserves disc structure and prevents post-discectomy disc degeneration. ...

Biomimetic Porous Ti6Al4V Implants: A Novel Interbody Fusion Cage via Gel‐Casting Technique to Promote Spine Fusion

... For balanced regionalization combining the two, the effectiveness varies with interaction intensity and semantic randomness. Real-world urban networks, a mix of strong and weak spatial constraints [71], and semantic distributions, a mix of highly functional and highly mixed areas [72], [73], provide conditions for balanced partitioning [74]. ...

MNCD-KE: a novel framework for simultaneous attribute- and interaction-based geographical regionalization
  • Citing Article
  • July 2024