The　growing　interest　in　ferroelectric　materials　has　witnessed　the　thriving　prospect　of　bio-inspired　artificial　neuromorphic　system,　where　multi-level　polarization　states　play　a　crucial　role.　In　this　work,　with　typical　BaTiO3　ferroelectric　thin　film　as　the　model　system,　we　explore　the　physical　effects　of　inhomogeneity　on　polarization　switching　dynamics　and　neuromorphic　performance.　Inhomogeneous　films　exhibited　pinched　polarization-electric　field　hysteresis　loops,　leading　to　a　high　recognition　accuracy　of　96.03%　for　hand-written　digits,　compared　to　about　10.31%　for　homogeneous　films.　The　inhomogeneity　in　switching　dynamics　was　analyzed　by　inhomogeneous　field　mechanism.　Diffusive　distributions　of　switching　time　and　local　electric　fields　were　observed,　aligning　with　experimental　results　and　the　expected　inhomogeneity.　The　prolonged　domain　wall　depinning　time　and　lowered　energy　consumption　suggest　the　potential　for　multi-level　polarization　states,　a　possibility　further　confirmed　by　phase-field　simulations　that　demonstrated　their　presence　during　long-term　potentiation/depression.　Our　work　highlights　the　positive　influence　of　inhomogeneity　in　enhancing　the　performance　of　ferroelectric-based　neuromorphic　systems.