ArchCAD-400K: A Large-Scale CAD drawings Dataset and New Baseline for Panoptic Symbol Spotting

The introduction of the ArchCAD-400K dataset marks a significant advancement in the field of architectural design and engineering. Comprising 413,062 chunks from 5,538 standardized CAD drawings, it is over 26 times larger than the previous largest dataset, which will greatly reduce the manual effort required for data annotation. This dataset not only enhances the diversity of drawings but also provides fine-grained annotations that are crucial for various applications. In conjunction with this dataset, the Dual-Pathway Symbol Spotter (DPSS) model has been developed, incorporating an adaptive fusion module that improves feature recognition. The DPSS has demonstrated state-of-the-art performance in panoptic symbol spotting, validating the effectiveness of the ArchCAD-400K dataset. Together, these innovations have the potential to drive significant advancements in architectural design and construction, paving the way for more efficient and accurate engineering applications.