Vat　photopolymerization　3D　printing　has　gained　significant　attention　due　to　its　fast　printing　speed　and　high　precision.　However,　the　absence　of　effective　quality　assurance　methods　greatly　limits　its　applications.　Current　vat　photopolymerization　cannot　deal　with　commonly　occurring　defects　during　the　printing　process,　such　as　random　bubbles　and　resin　underfill,　making　it　challenging　to　consistently　produce　products　that　match　designed　geometry　and　functions.　To　address　this,　we　propose　an　innovative　vat　photopolymerization　solution　via　visual-guided　in-situ　repair　to　effectively　eliminate　printing　defects.　By　utilizing　an　enhanced　YOLOv5　network　and　K-means　algorithm,　real-time　detection　of　bubbles　and　resin　underfill　can　be　achieved　using　image　analysis.　The　optimal　method　for　defect　repair　was　then　automatically　generated　via　the　adaptive　scraper　control　and　the　repair　slice　edge　smoothing　generation　algorithm,　which　was　immediately　received　by　hardware　to　adjust　the　ongoing　printing　parameters　without　interrupting　the　continuous　printing　process.　Experimental　results　demonstrate　that　the　aforementioned　strategy　can　accurately　differentiate　typical　defects　such　as　bubbles　and　resin　underfill,　and　precisely　carry　out　in-situ　defect　repairs.　Comparisons　between　repaired　samples　and　defect-free　samples　show　minimal　differences　in　surface　morphology　and　fracture　strength　(¡0.6%).　The　proposed　solution　is　also　applicable　to　various　precursors.　Clearly,　the　strategy　above　offers　a　universal　approach　to　avoid　defects　in　vat　photopolymerization,　which　effectively　enhances　printing　efficiency,　reduces　material　wastage,　and　ensures　the　quality,　accuracy,　and　reliability　of　printed　products.　©　2023　Elsevier　B.V.