FAQs¶
SDK version¶
View current SDK version
cd path-to-sdk/
realpath .repo/manifest.xml
Rollback SDK version
cd path-to-sdk/
# Link the SDK version file name that needs to be rolled back
ln -sf .repo/manifests/rv1126_rv1109/xxx.xml .repo/manifest.xml
# start back
repo sync -c
NPU version¶
SDK version is rv1126_rv1109_linux_release_20220324_v2.2.5b.xml version and date is 2022.03.24 or later NPU version defaults to NPU 1.7.0.
Switch NPU version: Modify the configuration file corresponding to buildroot according to the RK_CFG_BUILDROOT variable described in the mk file, and add the following configuration
BR2_PACKAGE_NPU_ARC=y
After compiling and burning the system, execute the /npu_arc/usr/bin/npu_arc_env.sh script on RV1126_RV1109 to configure the NPU 1.6.0 environment. The principle of this operation is to uninstall the 1.7.0 kernel NPU ko module driver, and then install the 1.6.0 kernel NPU ko module driver.
AI algorithm¶
ROCKX algorithm¶
Rock-X SDK is a set of AI component library based on RK3399Pro/RK180X platform. Developers can quickly build AI applications through the API interface provided by the Rock-X SDK. Currently available for RV1126_RV1109 platform. The ROCKX algorithm is free, but the disadvantage is that the interface function cannot be optimized. For more details about ROCKX, please refer to the sdk/external/rockx/doc/Rockchip_Developer_Guide_RockX_SDK_CN.pdf file in the SDK.
Arcsoft Face Recognition Algorithm¶
RV1126_RV1109 has been adapted to Arcsoft face recognition algorithm. Only applicable to NPU 1.6.0 version, not applicable to later NPU version (including NPU 1.7.0 version). Currently NPU 1.6.0 and above (excluding NPU 1.6.0) will cause serious memory leaks when using Arcsoft face recognition algorithm, please know. Arcsoft face algorithm activation code, please contact the business to obtain.
Salary train plate recognition algorithm¶
RV1126_RV1109 has been adapted to the train plate recognition algorithm. The official resource download interface can download the train plate recognition firmware. The license plate recognition application source code is located in the sdk/external/xhlpr_app folder in the SDK. For the license plate recognition algorithm activation code, please contact the business to obtain it.