English | 简体中文
Taking ResNet50_vd as an example, this document demonstrates how to convert classification model to RKNN model.
# Install paddle2onnx
pip install paddle2onnx
# Download ResNet50_vd model files and test images
wget https://bj.bcebos.com/paddlehub/fastdeploy/ResNet50_vd_infer.tgz
tar -xvf ResNet50_vd_infer.tgz
# From static map to ONNX model. Attention: Align the save_file with the zip file name
paddle2onnx --model_dir ResNet50_vd_infer \
--model_filename inference.pdmodel \
--params_filename inference.pdiparams \
--save_file ResNet50_vd_infer/ResNet50_vd_infer.onnx \
--enable_dev_version True \
--opset_version 10 \
--enable_onnx_checker True
# Fix shape. Attention: the inputs here should correspond to the name of the inputs shown in netron.app, which may be image or x
python -m paddle2onnx.optimize --input_model ResNet50_vd_infer/ResNet50_vd_infer.onnx \
--output_model ResNet50_vd_infer/ResNet50_vd_infer.onnx \
--input_shape_dict "{'inputs':[1,3,224,224]}"
Taking the example of RKNN model from RK3588, we need to edit tools/rknpu2/config/ResNet50_vd_infer_rknn.yaml to convert ONNX model to RKNN model.
If you need to perform the normalize operation on NPU, configure the normalize parameters based on your model. For example:
model_path: ./ResNet50_vd_infer/ResNet50_vd_infer.onnx
output_folder: ./ResNet50_vd_infer
mean:
-
- 123.675
- 116.28
- 103.53
std:
-
- 58.395
- 57.12
- 57.375
outputs_nodes:
do_quantization: False
dataset: "./ResNet50_vd_infer/dataset.txt"
To normalize on CPU, refer to the following yaml:
model_path: ./ResNet50_vd_infer/ResNet50_vd_infer.onnx
output_folder: ./ResNet50_vd_infer
mean:
-
- 0
- 0
- 0
std:
-
- 1
- 1
- 1
outputs_nodes:
do_quantization: False
dataset: "./ResNet50_vd_infer/dataset.txt"
Here we perform the normalize operation on NPU.
python tools/rknpu2/export.py \
--config_path tools/rknpu2/config/ResNet50_vd_infer_rknn.yaml \
--target_platform rk3588