Image　edge　detection　is　the　basis　for　precise　positioning　and　accurate　cutting　of　cloth　pattern　contours.　Compared　with　the　commonly　used　traditional　edge　detection　methods,　the　Holistically-nested　Edge　Detection　has　clearer　and　more　continuous　detection　results　including　the　reduction　of　the　false　detection　rate.　However,　this　method　extracts　a　coarser　thick　outline.　In　order　to　extract　a　high-precision　cloth　pattern　outline,　clearly　distinguish　the　main　body　of　the　pattern　from　the　background,　provide　convenience　for　the　follow-up　cutting　machine　for　accurate　cutting,　this　paper　proposes　an　improved　method　for　edge　detection　of　cloth　pattern　cutting　based　on　the　holistically-nested　Edge　Detection　method.　The　edge　refinement　and　smoothing　process　are　added,　where　the　edge　detection,　edge　refinement,　and　edge　smoothing　of　the　clothes　images　are　carried　out　in　sequences,　so　that　the　extracted　cloth　pattern　contour　is　continuous,　smooth,　and　detailed,　allowing　the　respect　of　the　cutting　requirements　of　the　cutting　machine　and　the　requirements　of　the　factory　production.　©　2021　IEEE.