Keeping walls straight: data model and training set size matter for deep learning in building generalization

Abstract

ABSTRACTDeep learning-backed models have shown their potential of conducting map generalization tasks. However, pioneering studies for raster-based building generalization encountered a common “wabbly-wall effect” that makes the predicted building shapes unrealistic. This effect was identified as a critical challenge in the existing studies. This work proposes a layered data representation model that separately stores a building for generalization and its context buildings in different channels. Incorporating adjustments to training sample generation and prediction tasks, we show how even without using more complex deep learning architectures, the widely used Residual U-Net can already produce straight walls for the generalized buildings and maintains rectangularity and parallelism of the buildings very well for building simplification and aggregation in the scale transition from 1:5,000 to 1:10,000 and 1:5,000 to 1:15,000, respectively. Experiments with visual evaluation and quantitative indicators such as Intersection over Union (IoU), fractality, and roughness index show that using a larger input tensor size is an easy but effective solution to improve prediction. Balancing samples with data augmentation and introducing an attention mechanism to increase network learning capacity can help in certain experiment settings but have obvious tradeoffs. In addition, we find that the defects observed in previous studies may be due to a lack of enough training samples. We thus conclude that the wabbly-wall challenge can be solved, paving the way for further studies of applying raster-based deep learning models on map generalization.POLICY HIGHLIGHTS Demonstrates the effectiveness of the proposed data structure with multiple evaluation indicatorsIdentifies a “wabbly-wall effect” a challenge in deep-learning backed image based map generalizationProposes a layered data structure that separates a target building and its surrounding buildings to ease the learning task in training deep learning models for raster-based map generalization.KEYWORDS: Map generalizationdeep learningrasterbuilding simplificationU-Net AcknowledgmentsThe authors also appreciate the comments of four anonymous reviewers which helped improve the paper.Disclosure statementNo potential conflict of interest was reported by the author(s).Data availability statementThe raw maps that support the findings are available by request to Dr Yu Feng (y.feng@tum.de). The codes for U-Net and its variants are from third-party authors who are not affiliated with this manuscript. The codes for data preprocessing and the models adapted from U-Net models are available here: https://doi.org/10.6084/m9.figshare.21901086.v1.Supplemental dataSupplemental data for this article can be accessed online at https://doi.org/10.1080/15230406.2023.2264757Notes1. https://github.com/LeeJunHyun/Image_SegmentationAdditional informationFundingThis research was supported by the Swiss National Science Foundation through project number [200021_204081] DeepGeneralization.