Image classifier using Tensor Flow Lite and Python.
Based on https://github.com/tensorflow/tensorflow/tree/master/tensorflow/lite/examples/python
Python 3.8.
Can be used without GPU.
-
Install Python
-
Update pip
pip install --upgrade pip
-
Install Tensor Flow (currently installing only Tensor Flow Lite not works for me)
pip install tensorflow
-
Install Pillow for loading image support
python -m pip install --upgrade Pillow
To run the classifier, type this: python label_image.py
If CUDA is not setupm Tensor Flow complains this, but the program works and quite fast:
... W tensorflow/stream_executor/platform/default/dso_loader.cc:60] Could not load dynamic library 'cudart64_110.dll'; dlerror: cudart64_110.dll not found
... I tensorflow/stream_executor/cuda/cudart_stub.cc:29] Ignore above cudart dlerror if you do not have a GPU set up on your machine.
0.919720: 653:military uniform
0.017762: 907:Windsor tie
0.007507: 668:mortarboard
0.005419: 466:bulletproof vest
0.003828: 458:bow tie, bow-tie, bowtie
time: 54.104ms
At the first run it takes some time to load Tensor Flow, but next it loads faster. Nethertheless for repetitive image processing consider looping inside the program.
For more Tensor Flow Lite models for image classification see here:
https://www.tensorflow.org/lite/models/image_classification/overview