In　the　industry　of　textile,　manual　visual　inspection　is　relied　by　traditional　defect　detection　to　complete　the　textile　quality　assessment,　which　is　inefficient　and　imprecise.　The　development　of　technology　on　image　processing　and　the　emergence　of　deep　learning　make　it　possible　for　machines　to　replace　human　eyes.　Computer　exploit　the　knowledge,　which　is　extracted　and　comprehended　from　image　of　object,　to　detect　and　control　in　the　actual　world　and　greatly　make　up　for　the　deficiency　of　the　traditional　manual　detection.　In　this　paper,　we　take　textile　gloves　as　research　object　and　identify　whether　they　contain　defects.　The　identification　for　textile　gloves　not　only　to　detect　the　surface　defects　as　conventional　textiles,　but　also　to　identify　the　defects　in　the　shape,　which　make　it　more　complex　compared　with　conventional　textile　detection.　In　order　to　realize　the　efficient　detection　of　the　defects　of　textile　gloves,　YOLOv3　target　detection　is　used　as　the　detection　method　to　identify　the　surface　defects　and　integra　gloves　of　textile　gloves　respectively.　In　the　process　of　identification,　the　integral　gloves　are　divided　into　two　categories:　shape　defect　or　not.　By　this　way,　the　defect　identification　of　textile　gloves　can　be　realized　comprehensively　from　two　perspectives:　local　detection　and　overall　classification.　Experiments　show　that　our　method　can　more　accurately　determine　whether　gloves　are　defective.　©　2022,　The　Author(s),　under　exclusive　license　to　Springer　Nature　Singapore　Pte　Ltd.