diff --git a/APP_Framework/Framework/knowing/kpu/k210_yolov2_detect_procedure/Readme.md b/APP_Framework/Framework/knowing/kpu/k210_yolov2_detect_procedure/Readme.md
new file mode 100644
index 000000000..7b04a9288
--- /dev/null
+++ b/APP_Framework/Framework/knowing/kpu/k210_yolov2_detect_procedure/Readme.md
@@ -0,0 +1,34 @@
+# YOLO Framework for K210 KPU
+
+## Usage
+
+* Include all src and header there in menuconfig, and copy codes like below in *main.c* in folder *Application*.
+* In board *edu-riscv64*, if all operations are right, the LCD should light and show the image from camera and YOLO inference.
+
+```C
+#include <stdio.h>
+#include <string.h>
+// #include <user_api.h>
+#include <transform.h>
+
+
+extern int FrameworkInit();
+extern void ApplicationOtaTaskInit(void);
+extern void k210_detect(char *json_file_path);
+
+int main(void)
+{
+	printf("Hello, world! \n");
+	FrameworkInit();
+#ifdef APPLICATION_OTA
+	ApplicationOtaTaskInit();
+#endif
+	k210_detect("instrusion.json");
+    return 0;
+}
+```
+
+## TODO
+
+* KPU drivers are still not completed, because it is undefined how to reisgter KPU's bus and device name in XIZI system. The framework is still directly using the SDK From Canaan Inc. But after the driver completed, it will be easy to adapt framework for the APIs.
+* Camera and LCD drivers between RT and XIZI are not completely compatible. So there are some marcos like `ifdef` to make the framework compatible for both systems.