diff --git a/sql/catalyst/src/main/scala/org/apache/spark/sql/catalyst/optimizer/Optimizer.scala b/sql/catalyst/src/main/scala/org/apache/spark/sql/catalyst/optimizer/Optimizer.scala
index a3ab89dc7114..7001fa12cc7a 100644
--- a/sql/catalyst/src/main/scala/org/apache/spark/sql/catalyst/optimizer/Optimizer.scala
+++ b/sql/catalyst/src/main/scala/org/apache/spark/sql/catalyst/optimizer/Optimizer.scala
@@ -160,18 +160,29 @@ object SamplePushDown extends Rule[LogicalPlan] {
  * representation of data item.  For example back to back map operations.
  */
 object EliminateSerialization extends Rule[LogicalPlan] {
-  def apply(plan: LogicalPlan): LogicalPlan = plan transform {
+  private def eliminateSerialization(p: LogicalPlan, isTop: Boolean): LogicalPlan = p transform {
     case d @ DeserializeToObject(_, _, s: SerializeFromObject)
         if d.outputObjectType == s.inputObjectType =>
       // A workaround for SPARK-14803. Remove this after it is fixed.
       if (d.outputObjectType.isInstanceOf[ObjectType] &&
           d.outputObjectType.asInstanceOf[ObjectType].cls == classOf[org.apache.spark.sql.Row]) {
         s.child
+      } else if (isTop) {
+        // If DeserializeToObject is at the top of logical plan, we don't need to preserve output
+        // expr id.
+        s.child
       } else {
         // Adds an extra Project here, to preserve the output expr id of `DeserializeToObject`.
         val objAttr = Alias(s.child.output.head, "obj")(exprId = d.output.head.exprId)
         Project(objAttr :: Nil, s.child)
       }
+  }
+
+  def apply(plan: LogicalPlan): LogicalPlan = plan transform {
+    case p if p.children.exists(_.isInstanceOf[DeserializeToObject]) =>
+      p.withNewChildren(p.children.map(eliminateSerialization(_, false)))
+    case d: DeserializeToObject =>
+      eliminateSerialization(d, true)
     case a @ AppendColumns(_, _, _, s: SerializeFromObject)
         if a.deserializer.dataType == s.inputObjectType =>
       AppendColumnsWithObject(a.func, s.serializer, a.serializer, s.child)
diff --git a/sql/catalyst/src/test/scala/org/apache/spark/sql/catalyst/optimizer/EliminateSerializationSuite.scala b/sql/catalyst/src/test/scala/org/apache/spark/sql/catalyst/optimizer/EliminateSerializationSuite.scala
index 3c033ddc374c..79835edf3f85 100644
--- a/sql/catalyst/src/test/scala/org/apache/spark/sql/catalyst/optimizer/EliminateSerializationSuite.scala
+++ b/sql/catalyst/src/test/scala/org/apache/spark/sql/catalyst/optimizer/EliminateSerializationSuite.scala
@@ -42,7 +42,7 @@ class EliminateSerializationSuite extends PlanTest {
     val input = LocalRelation('obj.obj(classOf[(Int, Int)]))
     val plan = input.serialize[(Int, Int)].deserialize[(Int, Int)].analyze
     val optimized = Optimize.execute(plan)
-    val expected = input.select('obj.as("obj")).analyze
+    val expected = input
     comparePlans(optimized, expected)
   }
 
diff --git a/sql/core/src/test/scala/org/apache/spark/sql/DatasetSuite.scala b/sql/core/src/test/scala/org/apache/spark/sql/DatasetSuite.scala
index 3cb4e52c6d41..0ae23e282b02 100644
--- a/sql/core/src/test/scala/org/apache/spark/sql/DatasetSuite.scala
+++ b/sql/core/src/test/scala/org/apache/spark/sql/DatasetSuite.scala
@@ -659,6 +659,13 @@ class DatasetSuite extends QueryTest with SharedSQLContext {
     checkDataset(DatasetTransform.addOne(dataset), 2, 3, 4)
   }
 
+  test("dataset.rdd with generic case class") {
+    val ds = Seq(Generic(1, 1.0), Generic(2, 2.0)).toDS
+    val ds2 = ds.map(g => Generic(g.id, g.value))
+    ds.rdd.map(r => r.id).count
+    ds2.rdd.map(r => r.id).count
+  }
+
   test("runtime null check for RowEncoder") {
     val schema = new StructType().add("i", IntegerType, nullable = false)
     val df = sqlContext.range(10).map(l => {
@@ -676,6 +683,8 @@ class DatasetSuite extends QueryTest with SharedSQLContext {
   }
 }
 
+case class Generic[T](id: T, value: Double)
+
 case class OtherTuple(_1: String, _2: Int)
 
 case class TupleClass(data: (Int, String))