Rewrite for review commentary

[reinvantveer]

• reply with new due date

• Include 'geometry' in the disambiguation table.

• "Tasks" consistently, instead of "cases"

• The three tasks followed by the sentence before the ":" sign does not seem to flow smoothly.

• Correct total count in table 2

• Create a line chart of the results in the results table to show that it doesn't clarify anything.

• Tasks: Simple count for neighborhoods (instead of population density) is enough. explain that simple inhabitants count is a simpler task than population density.

• Section 2.2 What is the complete source data set (5 million buildings), and what's the relationship between the sample dataset (160k) and the complete set?

• Create a figure map to show neighborhood, building and archaeological features

• Create a "related work" section

• Go through JGSY issues to find more relevant literature there and look at common article structure

• Better explain difference between CNN and LSTM. We want to include these models from the shallow family and these models from the deep family. Section 3.3 glosses over the selection of the LSTM a little too easily. LSTM is one of the most popular forms of RNN and Sketch-RNN uses it too.

• Create a 'related work' section. Better paper positioning within the body of scientific literature, such as

• • Zhu, X. X., Tuia, D., Mou, L., Xia, G. S., Zhang, L., Xu, F., & Fraundorfer, F. (2017). Deep Learning in Remote Sensing: A Comprehensive Review and List of Resources. IEEE Geoscience and Remote Sensing Magazine, 5(4), 8-36 and
• • Ball, J. E., Anderson, D. T., & Chan, C. S. (2017). Comprehensive survey of deep learning in remote sensing: theories, tools, and challenges for the community. Journal of Applied Remote Sensing, 11(04), 1 and
• • Cheng, G., Han, J. 2016. A Survey on Object Detection in Optical Remote Sensing Images, ISPRS Journal of Photogrammetry and Remote Sensing, 117: 11-28, 2016 and
• • Wang, Y., Zhang, L., Tong, X., Liu, S., & Fang, T. (2017). A feature extraction and similarity metric-learning framework for urban model retrieval. International Journal of Geographical Information Science, 31(9), 1749-1769 and
• • certainly Wu, Q., Diao, W., Dou, F., Sun, X., Zheng, X., Fu, K., & Zhao, F. (2016). Shape-based object extraction in high-resolution remote-sensing images using deep Boltzmann machine. International Journal of Remote Sensing, 37(24), 6012-6022.
• • T. Jiang, G.-S. Xia, and Q. Lu, “Sketch-based aerial image retrieval,” in
    IEEE International
    Conference on Image Processing (ICIP)
    , 2017.

• Clarify used coordinate system. P.5, Section3.1, longitude and latitude are bounded by the interval [-360, 360]? What coordinate system is used? Should it be [-180, 180]?

• Paragraph 1.1: add a list of either research questions or contributions that cannot be missed. Can we use deep learning methods for classification tasks on vector shapes?

• Add methodology section:

• • What are machine learning models,
• • Why are ML methods used in the first place?
• • methodology in machine learning are deep models as good as shallow models?
• • What are the tasks
• • How are the ML models compared? What is accuracy
• • What is train/val/test split?
• • What is a confusion matrix?

• Better explanation that the shallow models do not, but the deep models do work on geometries directly.

• More insight into model misclassification: create confusion matrices for all models, deep models single run.

• Section 3 becomes data or tasks. Requirements for enough examples, from different, real-world domains to explore the space for the different tasks as well as possible.

• • Neighborhoods
• • • Logistic regression
• • • Decision tree
• • • K-nearest neighbors
• • • SVM RBF
• • • CNN
• • • RNN
• • Buildings
• • • Logistic regression
• • • Decision tree
• • • K-nearest neighbors
• • • SVM RBF
• • • CNN
• • • RNN
• • Archaeology
• • • Logistic regression
• • • Decision tree
• • • K-nearest neighbors
• • • SVM RBF
• • • CNN
• • • RNN

• Clarify picture for figure 1.

• • Step one: add grid to geometry
• • Step two: represent as vector
• • Step three: normalize