DEEP LEARNING FOR URBAN PERCEPTION: REGRESSION-BASED COMPARING OF RESNET18, VGG19, AND EFFICIENTNET-B1

Year 2025, Volume: 26 Issue: 4, 399 - 415, 25.12.2025

Ali Ekincek , Saye Nihan Çabuk

https://doi.org/10.18038/estubtda.1721167

Abstract

Urban perception is a multidimensional phenomenon reflecting individuals’ evaluations of the urban environment and playing a critical role in planning and design processes aimed at improving quality of life. This study aims to predict six different themes of urban perception (beautiful, boring, depressing, lively, safe, wealthy) from street view images using regression-based deep learning methods. Three different deep learning architectures—ResNet18, VGG19, and EfficientNet-B1—were employed. The Place Pulse 2.0 dataset was utilized in the modeling process, with approximately 110,000 labeled street images processed through necessary preprocessing steps (resizing, cropping, tensor conversion, and normalization). Models were trained with an 80% training and 20% validation split. Performance evaluation was conducted using Mean Absolute Error (MAE), Root Mean Squared Error (RMSE), R2 and validation loss graphs. Findings indicate that the EfficientNet-B1 model achieved the lowest error values, particularly in the “safe” and “lively” themes, while the ResNet18 model offered more balanced and stable performance in terms of validation loss. The VGG19 model generally yielded higher error rates and exhibited a clear tendency toward overfitting. It was observed that theme-specific visual complexity directly affected model performance. In conclusion, while deep learning architectures prove effective in modeling urban perception through visual data, both the choice of architecture and the inherent nature of the theme play decisive roles in model performance. This study highlights the importance of architecture- and theme-sensitive model design in AI-supported analysis of urban perception.

Keywords

Urban perception , Deep learning , Regression , Street view imagery , Place Pulse

References

• [1] Lynch K. The Image of the City. Cambridge [Mass.] Technology Press, 1960.
• [2] Nasar JL. Urban design aesthetics: the evaluative qualities of building exteriors. Environ Behav, 1994; 26 (3), 377–401.
• [3] Muller E, Gemmell E, Choudhury I, Nathvani R, Metzler AB, Bennett J, Denton E, Flaxman S, Ezzati M. City-wide perceptions of neighbourhood quality using street view images. arXiv, 2022; 2211.12139.
• [4] Tang F, Zeng P, Wang L, Zhang L, Xu W. Urban perception evaluation and street refinement governance supported by street view visual elements analysis. Remote Sens, 2024; 16 (19), 3661.
• [5] Temurçin K, Keçeli K. Bir davranışsal coğrafya çalışması: Isparta şehri örneğinde uluslararası öğrencilerin kentsel mekân algısı. Süleyman Demirel Univ Fen-Edebiyat Fak Sos Bilim Derg, 2015; 36, 1–22.
• [6] Hu C, Jia S, Zhang F, Xiao C, Ruan M, Thrasher J, Li X. UPDExplainer: an interpretable transformer-based framework for urban physical disorder detection using street view imagery. ISPRS J Photogramm Remote Sens, 2023; 204, 209–222.
• [7] Biljecki F, Ito K. Street view imagery in urban analytics and GIS: a review. Landsc Urban Plan, 2021; 215, 104217.
• [8] Ma H, Wu D. A natural language processing-based approach: mapping human perception by understanding deep semantic features in street view images. arXiv, 2023; 2311.17354.
• [9] Dubey A, Naik N, Parikh D, Raskar R, Hidalgo CA. Deep learning the city: quantifying urban perception at a global scale. Comput Vis ECCV, 2016; 196–212.
• [10] Qiu W, Li W, Liu X, Huang X. Subjectively measured streetscape perceptions to inform urban design strategies for Shanghai. ISPRS Int J Geo-Inf, 2021; 10, 493.
• [11] Sangers R, van Gemert J, van Cranenburgh S. Explainability of deep learning models for urban space perception. arXiv, 2022; 2208.13555.
• [12] Naik N, Kominers S, Raskar R, Glaeser E, Hidalgo CA. The dynamics of physical urban change. 2016.
• [13] Naik N, Philipoom J, Raskar R, Hidalgo CA. Streetscore: predicting the perceived safety of one million streetscapes. IEEE Conf Comput Vis Pattern Recognit Workshops, 2014; 793–799.
• [14] Ordonez V, Berg TL. Learning high-level judgments of urban perception. Lect Notes Comput Sci, 2014; 8694, 494–510.
• [15] Salesses P, Schechtner K, Hidalgo CA. The collaborative image of the city: mapping the inequality of urban perception. PLoS One, 2013; 8, 68400.
• [16] De Nadai M, et al. Are safer looking neighborhoods more lively? a multimodal investigation into urban life. Proc ACM Int Conf Multimedia, 2016.
• [17] Harvey C, Aultman-Hall L. Measuring urban streetscapes for livability: a review of approaches. Prof Geogr, 2016; 68 (1), 149–158.
• [18] Naik NN. Visual urban sensing: understanding cities through computer vision. Massachusetts Institute of Technology, 2017.
• [19] Porzi L, Rota Bulò S, Lepri B, Ricci E. Predicting and understanding urban perception with convolutional neural networks. Proc ACM Int Conf Multimedia, 2015; 139–148.
• [20] Wang W, Yang S, He Z, Wang M, Zhang J, Zhang W. Urban perception of commercial activeness from satellite images and streetscapes. Web Conf Companion Proc, 2018; 647–654.
• [21] Zhang F, Zhang D, Liu Y, Lin H. Representing place locales using scene elements. Comput Environ Urban Syst, 2018; 71, 153–164.
• [22] Ito K, Biljecki F. Assessing bikeability with street view imagery and computer vision. Transp Res Part C Emerg Technol, 2021; 132, 103371.
• [23] Santos FA, Silva TH, Loureiro AAF, Villas LA. Automatic extraction of urban outdoor perception from geolocated free texts. Soc Netw Anal Min, 2020; 10, 1–23.
• [24] Burke M, Mbonambi S, Molala P, Sefala R. Rapid probabilistic interest learning from domain-specific pairwise image comparisons. arXiv, 2017; 1706.05850.
• [25] Tan M, Le QV. EfficientNet: improving accuracy and efficiency through AutoML and model scaling. arXiv, 2019; 1905.11946.
• [26] Naik N, Raskar R, Hidalgo CA. Cities are physical too: using computer vision to measure the quality and impact of urban appearance. Am Econ Rev, 2016; 106, 128–132.