Maneuvering target sensing will be an important service of future vehicular networks, where precise velocity estimation is one of the core tasks. To this end, the recently proposed integrated sensing and communications (ISAC) provides a promising platform for achieving accurate velocity estimation. However, with one mono-static ISAC base station (BS), only the radial projection of the true velocity can be estimated, which causes serious estimation error. In this paper, we investigate the estimation of the true velocity of a maneuvering target with the assistance of intelligent reflecting surfaces (IRSs). In particular, we propose an efficient velocity estimation algorithm exploiting the two perspectives from the BS and IRS to the target. In particular, we propose a two-stage scheme where the true velocity can be recovered based on the Doppler frequency of the BS-target link and BS-IRS-target link. Experimental results validate that the true velocity can be precisely recovered with the help of IRSs and demonstrate the advantage of adding the IRS.