Optimized Vectorizing of Building Structures with Switch: High-Efficiency Convolutional Channel-Switch Hybridization Strategy

TL;DR

This work tackles building footprint reconstruction as a planar graph problem, where standard CNNs incur heavy parameter overhead. It introduces Switch, a parameter-efficient shift operator that performs cross-channel XOR exchanges followed by a $1\times1$ fusion, and SwitchNN, which adds group-based parameter sharing to further compress capacity. Replacing all $3\times3$ convolutions in a U2Net backbone with Switch variants and coupling with an Edge Sampler for candidate-edge scoring, the method achieves competitive precision and F1 scores on SpaceNet while reducing memory usage and inference time. The contributions include the Switch and SwitchNN architectures, the Edge Sampler module, and comprehensive empirical validation against multiple baselines, demonstrating potential for real-time and mobile deployment in footprint reconstruction and related domains.

Abstract

The building planar graph reconstruction, a.k.a. footprint reconstruction, which lies in the domain of computer vision and geoinformatics, has been long afflicted with the challenge of redundant parameters in conventional convolutional models. Therefore, in this letter, we proposed an advanced and adaptive shift architecture, namely the Switch operator, which incorporates non-exponential growth parameters while retaining analogous functionalities to integrate local feature spatial information, resembling a high-dimensional convolution operation. The Switch operator, cross-channel operation, architecture implements the XOR operation to alternately exchange adjacent or diagonal features, and then blends alternating channels through a 1x1 convolution operation to consolidate information from different channels. The SwitchNN architecture, on the other hand, incorporates a group-based parameter-sharing mechanism inspired by the convolutional neural network process and thereby significantly reducing the number of parameters. We validated our proposed approach through experiments on the SpaceNet corpus, a publicly available dataset annotated with 2,001 buildings across the cities of Los Angeles, Las Vegas, and Paris. Our results demonstrate the effectiveness of this innovative architecture in building planar graph reconstruction from 2D building images.

Figures (3)

• Figure 1: Schematic illustration of Convolutional Neural Network (CNN) and Switch: (a) 3*3 Conv operation; (b) Switch operation; The 3*3 convolutional operation and the Switch operation have similar functionality with integrating local spatial features, as stated in (a) and (b). However, the number of parameters in the convolutional operation can increase exponentially with the number of channels, while the Switch operation can effectively reduce the parameters and facilitate fine-tuning of the model.
• Figure 2: Schematic illustration of Switch and SwitchNN, k is xor parameter, taken k = 5 as an example in here. (a) Switch is an activation function, which only contains a Switch operation without any parameter. (b) SwitchNN constructed a parameter of W, shared in the group, and combined 1x1 convolution to mix the channel's features. Interval Switch adopted in adjacent channels, i.e. odd and even for channels index.
• Figure 3: Building Planar Graph Reconstruction Results: (a) Ground Truth; (b) Nauata et al.; (c) Zhang et al.; (d) HEAT; (e) Roof-Former; (f) Our