Sci-Net: Scale Invariant Model for Buildings Segmentation from Aerial Imagery

TL;DR

Sci-Net is a novel scale-invariant neural network architecture designed for building segmentation in aerial images across diverse spatial resolutions, outperforming existing models on multiple datasets.

Contribution

Introduces Sci-Net, a scale-invariant model utilizing UNet and Dense ASPP to improve building segmentation across various image resolutions.

Findings

Outperforms state-of-the-art models on Open Cities AI dataset

Achieves steady improvement across different spatial resolutions

Effectively handles broad spectrum of aerial image resolutions

Abstract

Buildings' segmentation is a fundamental task in the field of earth observation and aerial imagery analysis. Most existing deep learning-based methods in the literature can be applied to a fixed or narrow-range spatial resolution imagery. In practical scenarios, users deal with a broad spectrum of image resolutions. Thus, a given aerial image often needs to be re-sampled to match the spatial resolution of the dataset used to train the deep learning model, which results in a degradation in segmentation performance. To overcome this challenge, we propose, in this manuscript, Scale-invariant Neural Network (Sci-Net) architecture that segments buildings from wide-range spatial resolution aerial images. Specifically, our approach leverages UNet hierarchical representation and Dense Atrous Spatial Pyramid Pooling to extract fine-grained multi-scale representations. Sci-Net significantly outperforms state of the art models on the Open Cities AI and the Multi-Scale Building datasets with a steady improvement margin across different spatial resolutions.