ResNet-50 Object Classification

Enter the Radxa Model-zoo ResNet directory
```
cd Radxa-Model-Zoo/sample/ResNet
```

Download the models, you can choose from F32/F16/INT8 quantized models

# F32
wget https://github.com/radxa-edge/TPU-Edge-AI/releases/download/model-zoo/resnet50_fp32_1b.bmodel

# F16
wget https://github.com/radxa-edge/TPU-Edge-AI/releases/download/model-zoo/resnet50_fp16_1b.bmodel

# INT8
wget https://github.com/radxa-edge/TPU-Edge-AI/releases/download/model-zoo/resnet50_int8_1b.bmodel

Download test images to the data folder

mkdir images && cd images
wget https://github.com/radxa-edge/TPU-Edge-AI/releases/download/model-zoo/grace_hopper.bmp

Create a virtual environment

It's necessary to create a virtual environment to avoid potential interference with other applications, for instructions on using virtual environments, please refer here
```
python3 -m virtualenv .venv
source .venv/bin/activate
```

Install python dependencies

pip3 install --upgrade pip
pip3 install numpy
pip3 install https://github.com/radxa-edge/TPU-Edge-AI/releases/download/v0.1.0/sophon_arm-3.7.0-py3-none-any.whl

Run the python examples

Under the python directory, a series of Python routines are provided, as follows:

No.	Python Example	Description
1	resnet_opencv.py	Decoding with OpenCV, preprocessing with OpenCV, inference with SAIL
2	resnet_bmcv.py	Decoding with SAIL, preprocessing with BMCV, inference with SAIL

Run resnet_opencv.py

export LD_LIBRARY_PATH=/opt/sophon/libsophon-current/lib:$LD_LIBRARY_PATH
export PYTHONPATH=$PYTHONPATH:/opt/sophon/sophon-opencv-latest/opencv-python/
python3 python/resnet_opencv.py --input ./images --bmodel ./resnet50_int8_1b.bmodel

Parameter explanation

resnet_opencv.py [--input IMG_PATH] [--bmodel BMODEL]

--input: path to the images for inference, the path to the folder containing the images can be input

--bmodel: path to the bmodel used for inference

(.venv) linaro@bm1684:/data/ssd/docs_check/Radxa-Model-Zoo/sample/ResNet$ python3 python/resnet_opencv.py --input ./images --bmodel ./resnet50_int8_1b.bmodel
[BMRT][bmcpu_setup:406] INFO:cpu_lib 'libcpuop.so' is loaded.
bmcpu init: skip cpu_user_defined
open usercpu.so, init user_cpu_init
[BMRT][load_bmodel:1084] INFO:Loading bmodel from [./resnet50_int8_1b.bmodel]. Thanks for your patience...
[BMRT][load_bmodel:1023] INFO:pre net num: 0, load net num: 1
INFO:root:filename: ./images/grace_hopper.bmp, res: (457, 0.12794505059719086)
INFO:root:result saved in ./results/resnet50_int8_1b.bmodel_dataset_opencv_python_result.json
INFO:root:------------------ Inference Time Info ----------------------
INFO:root:decode_time(ms): 2.94
INFO:root:preprocess_time(ms): 17.86
INFO:root:inference_time(ms): 4.36
INFO:root:postprocess_time(ms): 0.35

The results will be saved in ./results/resnet50_int8_1b.bmodel_dataset_opencv_python_result.json

[
  {
    "filename": "./dataset/grace_hopper.bmp",
    "prediction": 457,
    "score": 0.12794505059719086
  }
]

Run resnet_bmcv.py

export LD_LIBRARY_PATH=/opt/sophon/libsophon-current/lib:$LD_LIBRARY_PATH
export PYTHONPATH=$PYTHONPATH:/opt/sophon/sophon-opencv-latest/opencv-python/
python3 python/resnet_bmcv.py  --input ./images --bmodel ./resnet50_int8_1b.bmodel

Parameter explanation

resnet_bmcv.py [--input IMG_PATH] [--bmodel BMODEL]

--input: path to the images for inference, the path to the folder containing the images can be input

--bmodel: path to the bmodel used for inference

(.venv) linaro@bm1684:/data/ssd/docs_check/Radxa-Model-Zoo/sample/ResNet$ python3 python/resnet_bmcv.py  --input ./images --bmodel ./resnet50_int8_1b.bmodel
[BMRT][bmcpu_setup:406] INFO:cpu_lib 'libcpuop.so' is loaded.
bmcpu init: skip cpu_user_defined
open usercpu.so, init user_cpu_init
[BMRT][load_bmodel:1084] INFO:Loading bmodel from [./resnet50_int8_1b.bmodel]. Thanks for your patience...
[BMRT][load_bmodel:1023] INFO:pre net num: 0, load net num: 1
INFO:root:filename: ./images/grace_hopper.bmp, res: (457, 0.14356361329555511)
INFO:root:result saved in ./results/resnet50_int8_1b.bmodel_images_bmcv_python_result.json
INFO:root:------------------ Inference Time Info ----------------------
INFO:root:decode_time(ms): 4.01
INFO:root:preprocess_time(ms): 5.56
INFO:root:inference_time(ms): 1.86
INFO:root:postprocess_time(ms): 0.30

The results will be saved in ./results/resnet50_int8_1b.bmodel_images_bmcv_python_result.json

[
  {
    "filename": "./images/grace_hopper.bmp",
    "prediction": 457,
    "score": 0.14356361329555511
  }
]

FAQ

For more information on model conversion, model quantization details, and model accuracy testing, please refer to:

Model Conversion and Radxa Model-Zoo ResNet

ResNet-50 Object Classification

FAQ​

FAQ