Preserving human privacy in real estate listing applications by deep learning methods

Experimental results indicate that the 52 epoch ResNet-50 architecture is the best for our datasets with 98.45% overall accuracy and 98.00% precision, 98.90% recall, and 98.44% F1-score.

📌 3 Key Takeaways

🎯

Objective & Hypothesis:The images are important components of real estate applications on the internet to inform users.

📊

Key Findings & Data:The REP dataset was split into 70%, 20%, and 10% for training, validation, and testing, respectively.

💡

Conclusion & Impact:Experimental results indicate that the 52 epoch ResNet-50 architecture is the best for our datasets with 98.45% overall accuracy and 98.00% precision, 98.90% recall, and 98.44% F1-score.

🔬Design:Modeling & Simulation

📍Scope:Comprehensive Sample

⚙️Variables:Real Estate Privacy Resnet-50 Inceptionv3 DenseNet-169 · Real Estate · Privacy

Abstract

The images are important components of real estate applications on the internet to inform users. There are multiple rental and sale properties and many images of these properties on the internet, and it is challenging to control the images of these real estate in terms of time, workload, and cost. Considering the requirements of the problem, Deep Learning (DL), one of the Artificial Intelligence (AI) methods, offers ideal solutions. This study aims to distinguish images that contain humans using deep learning techniques. This will also aid in not violating the privacy of people according to the Law on the Protection of Personal Data in the image content used in real estate applications. For this purpose, firstly, a dataset of real estate images with and without humans called the Real Estate Privacy (REP) dataset was created. The REP dataset was split into 70%, 20%, and 10% for training, validation, and testing, respectively. Secondly, the REP dataset was trained with Inceptionv3, ResNet-50, and DenseNet-169 architectures using transfer learning. Lastly, the performances of the architectures were evaluated by accuracy, precision, recall, and F1-score accuracy metrics. Experimental results indicate that the 52 epoch ResNet-50 architecture is the best for our datasets with 98.45% overall accuracy and 98.00% precision, 98.90% recall, and 98.44% F1-score. The Inceptionv3 model provided the best results on the 55th epoch with 98.27% accuracy, 97.81% precision, 98.71% recall, and 98.26% F1-score. Finally, the DenseNet-169 model produced the best results on the 47th epoch, with 97.81% accuracy, 97.09% precision, 98.52% recall, and 97.80% F1-score. Accuracy assessment shows that the highest accuracy among the three architectures was obtained with the ResNet-50 architecture This study shows that deep learning methods offer a perspective to image content control and can be used efficiently in real estate applications.

Keywords

Real Estate Privacy Resnet-50 Inceptionv3 DenseNet-169 Real Estate Privacy ResNet-50 Inception v3 DenseNet-169

🔖

Cite This Article

VARUL, Y. E., ADIYAMAN, H., BAKIRMAN, T., BAYRAM, B., ALKAN, E., KARACA, S. Z., TOPALOĞLU, R. H. (2023). Preserving human privacy in real estate listing applications by deep learning methods. Mersin Photogrammetry Journal, Cilt: 5 Sayı: 1, 10–17. https://ajindex.com/articles/952683

📄 Full Text PDF Viewer

Open in New Tab ↗

↗New Tab

📄

PDF görüntüleyici hazırlanıyor...

References (10)

Fields, D., & Rogers, D. (2021). Towards a critical housing studies research agenda on platform real estate. Housing, theory and society, 38(1), 72-94.
Kumari, A., Maan, V., & Dhiraj. (2021). A Deep Learning-Based Segregation of Housing Image Data for Real Estate Application. In Intelligent Learning for Computer Vision: Proceedings of Congress on Intelligent Systems 2020 (pp. 165-179).
Springer Singapore. Bappy, J. H., Barr, J. R., Srinivasan, N., & Roy-Chowdhury, A. K. (2017, March). Real estate image classification. In 2017 ieee winter conference on applications of computer vision (wacv) (pp. 373-381).
IEEE. Azizi, I., & Rudnytskyi, I. (2022). Improving Real Estate Rental Estimations with Visual Data. Big Data and Cognitive Computing, 6(3), 96. Hassanzadeh, Z., Biddle, R., & Marsen, S. (2021). User perception of data breaches. IEEE Transactions on Professional Communication, 64(4), 374-389.
Erdos, D. (2022). Identification in personal data: Authenticating the meaning and reach of another broad concept in EU data protection law. Computer Law & Security Review, 46, 105721. Zhang, B., Zou, G., Qin, D., Ni, Q., Mao, H., & Li, M. (2022). RCL-Learning: ResNet and convolutional long short-term memory-based spatiotemporal air pollutant concentration prediction model. Expert Systems with Applications, 207, 118017. Salman, F. M., & Abu-Naser, S. S. (2022). Classification of real and fake human faces using deep learning. International Journal of Academic Engineering Research (IJAER), 6(3), 1-14
Krizhevsky, A., Sutskever, I., & Hinton, G. E. (2017). Imagenet classification with deep convolutional neural networks. Communications of the ACM, 60(6), 84-90.
Long, J., Shelhamer, E., & Darrell, T. (2015). Fully convolutional networks for semantic segmentation. In Proceedings of the IEEE conference on computer vision and pattern recognition (pp. 3431-3440).
Simonyan, K., & Zisserman, A. (2014). Very deep convolutional networks for large-scale image recognition. arXiv preprint arXiv:1409.1556. Szegedy, C., Liu, W., Jia, Y., Sermanet, P., Reed, S., Anguelov, D.,. & Rabinovich, A. (2015). Going deeper with convolutions. In Proceedings of the IEEE conference on computer vision and pattern recognition (pp. 1-9).
He, K., Zhang, X., Ren, S., & Sun, J. (2016). Deep residual learning for image recognition. In Proceedings of the IEEE conference on computer vision and pattern recognition (pp. 770-778).
Xie, S., Girshick, R., Dollár, P., Tu, Z., & He, K. (2017). Aggregated residual