Dataset of 30,000 video records for fine-tuning time-sensitive vision models.
benchmarkvideomultimodal

Notes

Date approximate.