The dataset contains vehicle trajectories, not detection annotations; those live in rfonod/songdo-vision. object-detection placed the card in a facet its users do not browse.
Fix Zenodo DOI: use concept DOI 13828383 instead of v1 DOI 13828384
10.5281/zenodo.13828384 is the v1 version DOI, not the concept DOI, so the badge, download instructions and citation all pointed at v1 (which lacks master_frames.zip). The concept DOI is 10.5281/zenodo.13828383, which resolves to the latest version (currently v2, record 17924857).
