NVIDIA Jetson Nano學習筆記（四）：安裝與執行Tensorflow Lite Model官方範例

3 min readMar 13, 2020

Python quickstart | TensorFlow Lite

Using TensorFlow Lite with Python is great for embedded devices based on Linux, such as Raspberry Pi and Coral devices…

www.tensorflow.org

更新

sudo apt-get update
sudo apt-get upgrade

安裝Cython或更新Cython

pip3 install Cython
pip3 install --upgrade Cython

安裝Numpy

sudo apt-get install python3-numpy
pip3 install numpy

安裝Tensorflow Lite(Python 3.6/ Linux ARM 64)

依照自己Nano的環境去調整Python 3.5 or 3.6 or 3.7。但是ARM 64 是肯定的。(Raspberry Pi為Linux ARM 32)

pip3 install https://dl.google.com/coral/python/tflite_runtime-2.1.0.post1-cp36-cp36m-linux_aarch64.whl

下載範例圖片

# 取得 photo(grace_hopper.bmp)
curl https://raw.githubusercontent.com/tensorflow/tensorflow/master/tensorflow/lite/examples/label_image/testdata/grace_hopper.bmp > /tmp/grace_hopper.bmp# 取得 model(mobilenet_v1_1.0_224.tflite)
curl https://storage.googleapis.com/download.tensorflow.org/models/mobilenet_v1_2018_02_22/mobilenet_v1_1.0_224.tgz | tar xzv -C /tmp# 取得labels(labels.txt)
curl https://storage.googleapis.com/download.tensorflow.org/models/mobilenet_v1_1.0_224_frozen.tgz  | tar xzv -C /tmp  mobilenet_v1_1.0_224/labels.txt# 將labels.txt移動到/tmp/
mv /tmp/mobilenet_v1_1.0_224/labels.txt /tmp/

執行官方範例

從以下指令可以看到，我們只需要將model_file、label_file、image，改成自己訓練的model、建立的label和想辨識的image，即可重製辨識結果。

python3 label_image.py \
  --model_file /tmp/mobilenet_v1_1.0_224.tflite \
  --label_file /tmp/labels.txt \
  --image /tmp/grace_hopper.bmp

輸出結果：

0.728693: military uniform
0.116163: Windsor tie
0.035517: bow tie
0.014874: mortarboard
0.011758: bolo tie