On 14 April 2010, a destructive earthquake (Ms 7.1) occurred along the Ganzi–Yushu fault (GYF), which is the boundary between the Songpan–Ganzi and Qiangtang Blocks of the eastern Tibetan Plateau. In this study, we present an effective workflow for microearthquake detection and location that assembles three main methods including a graphics processing unit-based match and locate technology (GPU M&L), a deep-neural-network-based noise-reduction technique (DeepDenoiser), and a U-net-based phase picking method (PhaseNet). Following this workflow, a high-resolution catalog has been constructed based on continuous seismic data from 2008 to 2018 recorded by 13 seismic stations deployed by Qinghai seismic network in Yushu, Qinghai. The GPU M&L is first applied to detect earthquakes which are usually hard to pick by routine methods due to their low signal-to-noise ratios. Then the DeepDenoiser is applied to remove noises from events we detected by GPU M&L, providing a better picking of seismic phases. The PhaseNet and hypoDD are conducted to pick seismic phases and relocate detected events, respectively. The spatial–temporal evolution of the foreshock sequence of the Yushu earthquake suggests that a triggered cascade model is responsible for the nucleation of the mainshock. The mainshock and aftershocks southeast of the mainshock mainly rupture on the primary northwest (NW)-trending GYF while the foreshocks and aftershocks on the NW segment rupture along the northeast (NE)-trending fault normal to the strike of the GYF. The average cumulative slip rate of the GYF estimated from six repeating earthquakes is <inline-formula xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink"> <tex-math notation="LaTeX">$7.4~\pm ~1.7$ </tex-math></inline-formula> mm/year, which is consistent with the geodetic and geological observations.