Beyond Segmentation: Structurally Informed Facade Parsing from Imperfect Images
Maciej Janicki, Aleksander Plocharski, Przemyslaw Musialski
TLDR
This paper improves facade parsing by adding a structural alignment loss to YOLOv8, making detections more coherent for reconstruction.
Key contributions
- Introduces a novel lightweight alignment loss for object detectors to enforce structural coherence in facade parsing.
- Integrates geometric priors into YOLOv8 training without modifying its standard inference pipeline.
- Significantly improves structural regularity and corrects alignment errors from perspective/occlusion.
- Achieves improved structural regularity on the CMP dataset with a controllable trade-off in detection accuracy.
Why it matters
Current facade parsing often lacks structural coherence, hindering downstream 3D reconstruction. This method injects structural intelligence into existing detectors, making outputs more useful for architectural modeling and urban planning. It bridges a critical gap between detection and reconstruction.
Original Abstract
Standard object detectors typically treat architectural elements independently, often resulting in facade parsings that lack the structural coherence required for downstream procedural reconstruction. We address this limitation by augmenting the YOLOv8 training objective with a custom lightweight alignment loss. This regularization encourages grid-consistent arrangements of bounding boxes during training, effectively injecting geometric priors without altering the standard inference pipeline. Experiments on the CMP dataset demonstrate that our method successfully improves structural regularity, correcting alignment errors caused by perspective and occlusion while maintaining a controllable trade-off with standard detection accuracy.
📬 Weekly AI Paper Digest
Get the top 10 AI/ML arXiv papers from the week — summarized, scored, and delivered to your inbox every Monday.