Seismic tomography is one of the most important tools that probe the Earth's interior and provide direct evidence to geodynamic processes of the mantle. It is also mathematically an inverse problem, with recordings of ground motion (i.e., seismograms) as the input data, and the 3D elastic and anelastic structures as model parameters. In this talk, I will summarize the recent advances in seismic imaging, transitioning from simple ray-based tomographic methods to full waveform inversions (FWI, also known as adjoint tomography in earthquake seismology). These FWI techniques rely on advanced forward numerical solvers, calculate accurate sensitivity kernels (i.e., gradient) based on adjoint methods, and iteratively improve velocity models by local optimization algorithms. Applications of FWIs to seismic imaging in southern California, Europe, East Asia and Australia have revealed high-resolution seismic features that significantly improved our understanding of tectonics at these regions.