diff --git a/pcdet/ops/pointnet2/pointnet2_batch/src/ball_query_gpu.h b/pcdet/ops/pointnet2/pointnet2_batch/src/ball_query_gpu.h
new file mode 100644
index 0000000..1213dda
--- /dev/null
+++ b/pcdet/ops/pointnet2/pointnet2_batch/src/ball_query_gpu.h
@@ -0,0 +1,15 @@
+#ifndef _BALL_QUERY_GPU_H
+#define _BALL_QUERY_GPU_H
+
+#include <torch/serialize/tensor.h>
+#include <vector>
+#include <cuda.h>
+#include <cuda_runtime_api.h>
+
+int ball_query_wrapper_fast(int b, int n, int m, float radius, int nsample, 
+	at::Tensor new_xyz_tensor, at::Tensor xyz_tensor, at::Tensor idx_tensor);
+
+void ball_query_kernel_launcher_fast(int b, int n, int m, float radius, int nsample, 
+	const float *xyz, const float *new_xyz, int *idx);
+
+#endif