Recognition　and　localization　of　the　fittings　in　overhead　transmission　line　(OTL)　images　are　the　basis　of　fittings　status　detection　and　fault　diagnosis.　Most　of　the　current　recognition　and　localization　of　the　fittings　are　based　on　object　detection　methods,　which　cannot　accurately　locate　the　fittings,　and　existing　instance　segmentation　methods　have　limited　accuracy　for　instance　segmentation　of　fittings　in　complex　OTL　scenarios.　To　solve　these　problems,　in　the　interleaved　execution　part,　we　inherit　the　idea　of　Hybrid　Task　Cascade　(HTC)　and　add　a　direct　information　path　to　the　same-stage　box　and　mask　branches　to　reinforce　the　coupled　relationship　between　detection　and　segmentation;　in　the　mask　branch　part,　we　sequentially　apply　the　efficient　channel　attention　(ECA)　module　and　the　dilated　spatial　attention　(DSA)　module　and　then　insert　them　into　the　mask　branch　to　improve　the　cross-stage　information　communication　in　the　cascade　architecture　and　mask　prediction.　Combining　them　results　in　Mixed　Attention　Interleaved　Execution　Cascade　(MAIEC),　a　new　cascade　architecture　for　instance　segmentation.　Extensive　experiments　on　the　OTL　fittings　dataset　reveal　the　effectiveness　of　the　proposed　method.　The　proposed　MAIEC　improves　the　AP　of　the　box　and　mask　predictions　by　respectively　2.0%　and　1.5%　compared　to　the　strong　HTC　baseline.　©　2023　IEEE.