Hi,
We are using imx95 develop some machine learning app. so I need some sample code and doc to check npu performance and learn how to use the npu to calculate complex operation.
first question:
I reference the document from nxp, it says we can use "label_image" to check the performance on npu and cpu.
- npu version: ./label_image -m mobilenet_v1_1.0_224_quant.tflite -i grace_hopper.bmp -l labels.txt --external_delegate_path=/usr/lib/libneutron_delegate.so
- cpu version: ./label_image -m mobilenet_v1_1.0_224_quant.tflite -i grace_hopper.bmp -l labels.txt .
the average time always return 16.xxx ms, why it doesn't improve when i using npu.
INFO: Loaded model mobilenet_v1_1.0_224_quant.tflite
INFO: resolved reporter
INFO: invoked
INFO: average time: 16.558 ms
INFO: 0.764706: 653 military uniform
INFO: 0.121569: 907 Windsor tie
INFO: 0.0156863: 458 bow tie
INFO: 0.0117647: 466 bulletproof vest
INFO: 0.00784314: 835 suit
the second question is i see /usr/lib/libneutron_delegate.so can load "NeutronKernels.bin and NeutronFirmware.elf", can we get the source code?
reference document
- https://www.nxp.com/docs/en/user-guide/IMX-MACHINE-LEARNING-UG.pdf
- https://github.com/nxp-imx/neutron
- https://github.com/nxp-imx/tflite-neutron-delegate
