A novel computer vision-based approach for autonomous building facade material stock estimation

Le résumé fourni par la source

Estimating building stock at the elemental level is essential for a circular economy. However, the lack of information on building stock due to considerable time and costs associated with data collection poses a significant challenge in large-scale analysis. This study employed computer vision-based façade classification and weighted material intensity to estimate building façade material stock using a bottom-up approach. Autonomous semantic enhancements on façade material typologies and opening areas aided better estimation of building-facing materials’ weight. ResNet 50 deep learning architecture was chosen, with an F1 score of 0.77 and 80 % accuracy. Most façade classes have achieved a relatively high level of accuracy of predictions. “Common Brick”, “Face Brick”, “Concrete Blocks”, and “Concrete” were identified as the most widespread façade materials in the model deployed area of interest: Southport (SA3), City of Gold Coast, Australia. The model provides the necessary foundations for progressing circular economy and urban metabolism efforts.

Ce résumé expose les affirmations des auteurs. BNTIC ne l’interprète pas comme une validation indépendante des résultats.