Improve performance by shrinking the vgg_faces.pb file [16h]

elimu-ai / ml-authentication

😶‍🌫️ Android application for authenticating the current student using face recognition

MIT License

16 stars 16 forks source link

Improve performance by shrinking the vgg_faces.pb file [16h] #391

Open sladomic opened 7 years ago

sladomic commented 7 years ago

[x] https://www.tensorflow.org/performance/quantization

bazel build tensorflow/tools/quantization/tools:quantize_graph
bazel-bin/tensorflow/tools/quantization/quantize_graph --input=tmp/vgg_faces.pb --output_node_names="fc7/fc7" --output=tmp/quantized_vgg_faces.pb --mode=eightbit

[ ] https://www.tensorflow.org/performance/xla/tfcompile