Questions regarding testing /workspace/examples/LittleNet/LittleNet.onnx model on KL720

William

Hi,

I have already created NEF file for Kl720 using /workspace/examples/LittleNet/LittleNet.onnx in the toolchain environment. Now I want to test the model on kl720.For this I want to write an inferencing code so I have a few questions

1) Do I need to use KDP wrapper , if yes what all APIs are needed

2) What is the difference between ISI mode and DME mode . Which mode to use for /workspace/examples/LittleNet/LittleNet.onnx model

3) What values should in use in model_id and image_cfg which is needed for kdp_wrapper.init_dme_config ? The model is /workspace/examples/LittleNet/LittleNet.onnx and the images used is /workspace/examples/LittleNet/pytorch_imgs/Abdullah_0001.png

/workspace/examples/LittleNet/pytorch_imgs/Abdullah_0002.png

/workspace/examples/LittleNet/pytorch_imgs/Abdullah_0003.png

/workspace/examples/LittleNet/pytorch_imgs/Abdullah_0004.png

I had already posted some of the questions in kl720 forum but didn't get any response so posting it here.

Please answer my queries

PowHuang

I post a comment for littlenet test at the 720 forunm .

The littlenet is just a example for toolchain model convert.

Yolo example is a more better to learn kneron APIs.

1. kdp_wrapper is necessary for python example
2. 720 not support dme mode now, 720 use isi_load_model to replace dme. Anyway ,use isi_load_model to put model in ddr, or use update_nef_model to put model in Flash. Two way could work for model inference.