To understand the deep structure and seismogenesis in the north-south seismic zone (NSSZ) in southwest China, we collected a large number of P and S wave arrival time data from both local earthquakes and teleseismic events to invert simultaneously for the three-dimensional P and S wave velocity (Vp, Vs) structures in the crust and upper mantle. The obtained Vp and Vs models are generally coherent with each other from the crust to the mantle transition zone. The NSSZ is characterized by strong structural heterogeneities from the mountainous areas of southeastern Tibet to the Yangtze Platform. Low-Vp and low-Vs anomalies are generally imaged in the upper crust under the Sichuan basin, showing good agreement with the surface geological features that the foreland basin contains primarily Mesozoic and Paleozoic sedimentary rocks with a thickness of several kilometers. The deep part of the Yangtze Platform is characterized by a cratonic lithospheric body dipping southwestward down to 400 km depth beneath the southeastern margin of the Tibetan Plateau, in sharp contrast to low-velocity anomalies in the upper mantle beneath the Songpan-Ganzi Fold System and the Northwest Qiangtang Block. At depths of the lower crust and uppermost mantle, a layer with low Vp and low Vs (1%-2%) is revealed under a region west of the Yangtze Platform, which may reflect the ductile flow in the lower crust. Our results indicate that the seismotectonics in the NSSZ is affected significantly by the ductile flow in the lower crust and strong heterogeneities in the upper mantle under southwest China in addition to the dominant role of the India-Asia collision.