Run the test program using rknn-toolkit-lite2 for rk3568 development board

1、 First of all, you need to deploy the operating environment , Put the library file in RK3568 Development board , We will send the online data “iTOP-3568 Development board \02_【iTOP-RK3568 Development board 】 Development of information \11_NPU Use supporting materials \05_Linux_librknn_api\librknn_api\aarch64” Files under path pass U Copy the disk to the development board /usr/lib Catalog , As shown in the figure below ：

2、 Download the network disk data “iTOP-3568 Development board \02_【iTOP-RK3568 Development board 】 Development of information \11_NPU Use supporting materials \04_RKNN-Toolkit-lite2 Tools ” in rknn_toolkit_lite2.tar.gz Compressed package , stay windows Decompress , And then through U Disk copy rknn_toolkit_lite2/examples Under folder Demo inference_with_lite Put in RK3568 Development board , This demo It's through resnet Residual neural network identification ”space_shuttle_224.jpg” This picture , And then run , Enter the following command ：

cd inference_with_lite/

python3 test.py

Run successfully, as shown in the figure below ：

Pass the result TOP The training set tag query of , The recognition result is a spaceship , Training set labels can be found in

The confidence is 0.9996, Is the highest of all labels , It is consistent with the picture .

More about ： Beijing Xunwei