13CO clump catalog from FUGIN in the Milky Way

First, the source extraction code FacetClumps was used to detect as many molecular clump candidates as possible from the FUGIN ^13^CO data. Second, a trained and validated semi-supervised deep learning model, SS-3D-Clump, was applied to verify these candidates, providing confidence levels for the clumps and filtering out false candidates to enhance the accuracy of the detection results. The resulting catalog containing 23150 clumps extracted from the ^13^CO (J=1-0) data covers the first quadrant (10<=l<=50{deg}, |b|<=1{deg}).

By matching with CHIMPS and inheriting the distances of the matched CHIMPS clumps, we found that the sizes of the FUGIN clumps range from 0.1 to 3pc, demonstrating that the dense structures belong to the clump scale. The catalog achieves an 80% completeness level above 466K.km/s. The proposed two-step approach effectively integrates clump detection algorithms with semi-supervised deep learning, achieving an accuracy comparable to manual verification and thereby improving the extraction of clumps from large-scale survey data.

The resulting clump catalog enables the analysis of the physical properties of clumps in high-mass star-forming regions, contributing to a better understanding of environmental influences on clump formation and the star formation process.