Need to get case code for pre trained onnx model SDK

VIM3
1

I have create the case code for above file but output is incorrect i couldn’t find what went wrong .i am trying to print layer wise information.
Need to know what parameter need to change for different framework (tf ,pytorch ,onnx) to convert pre-trained models
TF pretrained model casecodes are working fine but when i tried with onnx the accuracy is not right
Import
$convert_onnx
–onnx-model ./onnx_models/${NAME}.onnx
–net-output ${NAME}.json
–data-output ${NAME}.data
Quantize$tensorzone
–action quantization
–dtype float32
–source text
–source-file data/validation_tf.txt
–model-input ${NAME}.json
–model-data ${NAME}.data
–quantized-dtype asymmetric_affine-u8
–quantized-rebuild \

Export
$export_ovxlib
–model-input ${NAME}.json
–data-input ${NAME}.data
–model-quantize ${NAME}.quantize
–export-dtype quantized \
#–optimize VIPNANOQI_PID0X88
#–viv-sdk ${ACUITY_PATH}vcmdtools
#–pack-nbg-unify \

2

@Omkar_Shende How did you generate the executable code, and how did you do the post-processing, you only told me that the result was wrong, I couldn’t tell where the problem was

3

After getting the case code , i copied the build_vx, makefile.linux and makefile.linux.def to the case code then send the cas code to khadas board .Executed build_vx command which created me bin_r_c4
inside it there was executable.
when i ran the command for ./detect_model …/export.data file ~/image_path
the output was not accurate
Output
— Top5 —
103: 0.024506
885: 0.024506
977: 0.024506
6: 0.023254
58: 0.022064

$echo ‘TARGET_NAME = detect_{model_name}’ > makefile.target_name
wget -nc https://raw.githubusercontent.com/khadas/aml_npu_app/master/detect_library/inception/makefile.linux.def
wget -nc https://raw.githubusercontent.com/khadas/aml_npu_app/master/detect_library/inception/build_vx.sh