In　this　study,　taking　the　common　diseases　in　tomato　leaves,　which　are　typical　crops　in　southern　China,　as　the　research　object,　a　FC-SNDPN　(Fully　Convolutional　–　Switchable　Normalization　Dual　Path　Networks)　-based　method　for　automatic　identification　and　detection　of　crop　leaf　diseases　is　proposed　to　solve　the　problem　that　traditional　image　identification　methods　for　crop　diseases　and　insect　pests　heavily　rely　on　artificial　feature　extraction　and　have　a　poor　generalization　ability　for　image　recognition　with　a　complex　background.　In　order　to　reduce　the　influence　of　the　complicated　background　on　the　recognition　of　crops　diseases　and　insect　pests　image,　A　full　convolutional　network　(FCN)　algorithm　based　on　VGG-16　model　is　used　to　segment　the　target　crop　image.　Then　an　improved　DPN　(Dual-Path　Networks)　model　is　proposed　to　improve　the　ability　of　feature　extraction.　SNDPN　combines　the　connection　method　between　Desnet　and　Resnet　layers,　forms　a　neural　network　by　using　SN　layer,　and　adaptively　optimizes　the　parameters　of　the　dual-path　neural　network　by　switching　the　normalized　layer,　which　improves　the　versatility　of　the　network　for　different　types　of　diseases　and　insect　pests　and　the　training　speed　of　the　network.　Finally,　the　identification　accuracy　of　the　proposed　method　of　using　FCN　for　foreground　segmentation　and　SNDPN　for　identification　is　97.59%　on　the　augmentation　data　set,　the　result　proves　the　effectiveness　of　our　method.　©　2022,　The　Author(s),　under　exclusive　licence　to　Springer　Science+Business　Media,　LLC,　part　of　Springer　Nature.