Roof Material Classification from Aerial Imagery
Tóm tắt
this paper describes an algorithm for classification of roof materials using aerial photographs. Main advantages of the algorithm are proposed methods to improve prediction accuracy. Proposed methods includes: method of converting ImageNet weights of neural networks for using multi-channel images; special set of features of second level models that are used in addition to specific predictions of neural networks; special set of image augmentations that improve training accuracy. In addition, complete flow for solving this problem is proposed. The following content is available in open access: solution code, weight sets and architecture of the used neural networks. The proposed solution achieved second place in the competition “Open AI Caribbean Challenge”.
Tài liệu tham khảo
Wahba, S., Triveno, L., et al., Building better before the next disaster: How retrofitting homes can save lives and strengthen economies, 2018. https://blogs.worldbank.org/sustainablecities/building-better-next-disaster-how-retrofitting-homes-can-save-lives-and-strengthen-economies.
Open AI Caribbean challenge: mapping disaster risk from aerial imagery. https://www.drivendata.org/competitions/58/disaster-response-roof-type/.
Rawat, W. and Wang, Z., Deep convolutional neural networks for image classification: a comprehensive review, Neural Comput., 2017, vol. 29, no. 9, pp. 2352–2449.
Solovyev, R.A., Stempkovsky, A.L., and Telpukhov, D.V., Study of fault tolerance methods for hardware implementations of convolutional neural networks, Opt. Mem. Neural Networks, 2019, vol. 28, no. 2, pp. 82–88.
Yu, S., Jia, S., and Xu, C., Convolutional neural networks for hyperspectral image classification, Neurocomputing, 2017, vol. 219, pp. 88–98.
Iglovikov, V., et al., TernausNetV2: fully convolutional network for instance segmentation, Proc. 2018 IEEE/CVF Conf. on Computer Vision and Pattern Recognition Workshops (CVPRW), Piscataway, NJ: Inst. Electr. Electron. Eng., 2018, vol. 233, p. 237.
Soloviev, R.A., Telpukhov, D.V., and Kustov, A.G., Automatic segmentation of satellite images based on modified convolutional neural network UNET, Inzh. Vestn. Dona, 2017, vol. 47, no. 4 (47).
Buslaev, A., et al., Fully convolutional network for automatic road extraction from satellite imagery, Proc. 2018 IEEE/CVF Conf. on Computer Vision and Pattern Recognition Workshops (CVPRW), Piscataway, NJ: Inst. Electr. Electron. Eng., 2018, pp. 207–210.
Seferbekov, S.S., et al., Feature pyramid network for multi-class land segmentation, Proc. 2018 IEEE/CVF Conf. on Computer Vision and Pattern Recognition Workshops (CVPRW), Piscataway, NJ: Inst. Electr. Electron. Eng., 2018, pp. 272–275.
Zhang, C., et al., Automatic identification of center pivot irrigation systems from Landsat images using convolutional neural networks, Agriculture, 2018, vol. 8, no. 10, p. 147.
Deng, J., et al., A large-scale hierarchical image database, Proc. 2009 IEEE Conf. on Computer Vision and Pattern Recognition, Piscataway, NJ: Inst. Electr. Electron. Eng., 2009, pp. 248–255.
He, K., et al., Deep residual learning for image recognition, Proc. 2016 IEEE Conf. on Computer Vision and Pattern Recognition, Piscataway, NJ: Inst. Electr. Electron. Eng., 2016, pp. 770–778.
Huang, G., et al., Densely connected convolutional networks, Proc. 2017 IEEE Conf. on Computer Vision and Pattern Recognition, Piscataway, NJ: Inst. Electr. Electron. Eng., 2017, pp. 4700–4708.
Tan, M. and Le, Q.V., EfficientNet: rethinking model scaling for convolutional neural networks, 2019. https://arxiv.org/abs/1905.11946.
Sandler, M., et al., Mobilenetv2: Inverted residuals and linear bottlenecks, Proc. 2018 IEEE Conf. on Computer Vision and Pattern Recognition, Piscataway, NJ: Inst. Electr. Electron. Eng., 2018, pp. 4510–4520.
Brownlee, J., How to use out-of-fold predictions in machine learning, 2019. https://machinelearningmastery.com/out-of-fold-predictions-in-machine-learning/.
Simard, P.Y., et al., Best practices for convolutional neural networks applied to visual document analysis, Proc. Seventh Int. Conf. on Document Analysis and Recognition, (ICDAR 2003), Piscataway, NJ: Inst. Electr. Electron. Eng., 2003, vol. 3.
Wigington, C., et al., Data augmentation for recognition of handwritten words and lines using a CNN-LSTM network, Proc. 14th IAPR Int. Conf. on Document Analysis and Recognition (ICDAR), Piscataway, NJ: Inst. Electr. Electron. Eng., 2017, vol. 1, pp. 639–645.
Fitzgibbon, A.W., Simultaneous linear estimation of multiple view geometry and lens distortion, Proc. 2001 IEEE Computer Society Conf. on Computer Vision and Pattern Recognition (CVPR 2001), Piscataway, NJ: Inst. Electr. Electron. Eng., 2001, vol. 1, pp. 125–132.
Szegedy, C., et al., Inception-v4, inception-resnet and the impact of residual connections on learning, Proc. Thirty-First AAAI Conf. on Artificial Intelligence, Menlo Park, CA: Assoc. Adv. Artif. Intell., 2017.
Hu, J., Shen, L., and Sun, G. Squeeze-and-excitation networks, Proc. 2018 IEEE Conf. on Computer Vision and Pattern Recognition, Piscataway, NJ: Inst. Electr. Electron. Eng., 2018, pp. 7132–7141.
Chen, T. and Guestrin, C., Xgboost: a scalable tree boosting system, Proc. 22nd ACM SIGKDD Int. Conf. on Knowledge Discovery and Data Mining, New York: Assoc. Comput. Mach., 2016, pp. 785–794.
Ke, G., et al., LightGBM: A highly efficient gradient boosting decision tree, Proc. 31st Int. Conf. on Neural Information Processing Systems (NIPS'17), Red Hook, NY: Curran Assoc., 2017, pp. 3146–3154.
Prokhorenkova, L., et al., CatBoost: unbiased boosting with categorical features, Proc. 32nd Int. Conf. on Neural Information Processing Systems (NIPS'18), Red Hook, NY: Curran Assoc., 2018, pp. 6638–6648.
GitHub website. https://github.com/ZFTurbo/DrivenData-Open-AI-Caribbean-Challenge-2nd-place-solution.