diff --git a/lite/kernels/host/multiclass_nms_compute.cc b/lite/kernels/host/multiclass_nms_compute.cc
index 9cbc798d46ecb3cf98159e9b4762c8692ec8c1eb..9f4c2fb6f53c17f97cb9a1aaeecd493713899cab 100644
--- a/lite/kernels/host/multiclass_nms_compute.cc
+++ b/lite/kernels/host/multiclass_nms_compute.cc
@@ -429,6 +429,17 @@ REGISTER_LITE_KERNEL(multiclass_nms,
     .BindInput("BBoxes", {LiteType::GetTensorTy(TARGET(kHost))})
     .BindInput("Scores", {LiteType::GetTensorTy(TARGET(kHost))})
     .BindOutput("Out", {LiteType::GetTensorTy(TARGET(kHost))})
+    .Finalize();
+
+REGISTER_LITE_KERNEL(multiclass_nms2,
+                     kHost,
+                     kFloat,
+                     kNCHW,
+                     paddle::lite::kernels::host::MulticlassNmsCompute,
+                     def)
+    .BindInput("BBoxes", {LiteType::GetTensorTy(TARGET(kHost))})
+    .BindInput("Scores", {LiteType::GetTensorTy(TARGET(kHost))})
+    .BindOutput("Out", {LiteType::GetTensorTy(TARGET(kHost))})
     .BindOutput("Index",
                 {LiteType::GetTensorTy(TARGET(kHost), PRECISION(kInt32))})
     .Finalize();
diff --git a/lite/operators/multiclass_nms_op.cc b/lite/operators/multiclass_nms_op.cc
index 9dba5de4f81a1cba8f66132d89f6321ed76d368c..9ec79f8b57d63f20325bf686c1280522aa4fa80a 100644
--- a/lite/operators/multiclass_nms_op.cc
+++ b/lite/operators/multiclass_nms_op.cc
@@ -84,3 +84,4 @@ bool MulticlassNmsOpLite::AttachImpl(const cpp::OpDesc& opdesc,
 }  // namespace paddle
 
 REGISTER_LITE_OP(multiclass_nms, paddle::lite::operators::MulticlassNmsOpLite);
+REGISTER_LITE_OP(multiclass_nms2, paddle::lite::operators::MulticlassNmsOpLite);