Flower　classification　is　an　important　branch　in　the　field　of　plant　classification.　Flower　images　are　difficult　to　improve　recognition　accuracy　due　to　the　similarity　of.　In　recent　years,　with　the　continuous　development　of　deep　learning　convolutional　neural　networks,　convolutional　neural　networks　have　been　increasingly　used　in　flower　image　classification.　However　convolutional　neural　networks　have　a　large　number　of　model　parameters,　parameter　redundancy,　and　are　difficult　to　deploy　on　mobile　devices.　To　solve　the　above　problems,　this　paper　the　following　work:　1.　Training　the　flower　dataset　on　the　convolutional　neural　network　model　VGG16.　2.　Using　transfer　learning,　the　results　obtained　training　are　trained　on　the　pruned　model.　3.　Setting　different　pruning　rates,　retraining　the　model　to　get　the　final　result.　Experiments　show　that　using　pruning　algorithm　to　train　the　flower　dataset　improves　accuracy,　with　the　highest　accuracy　being　93.10%.　©　COPYRIGHT　SPIE.