Efficientnet-EdgeTPU are a family of image classification neural network models customized for deployment on Google Edge TPU. These networks are closely related to Efficientnets.
引用
Efficientnet-EdgeTPU were developed using the AutoML MNAS framework by augmenting the neural network search
space with building blocks tuned to execute efficiently on the EdgeTPU neural network accelerator architecture.
The neural architecture search was incentivized to discover models that achieve low parameter footprint
and low latency on EdgeTpu, while simultaneously achieving high classification accuracy.
This neural architecture search produced a baseline model: edgetpunet-S,
which is subsequently scaled up using EfficientNet's compound scaling method to produce the M and L models.
とありますね。。。AutoML MNAS frameworkで開発されたんですね。。。
ResNet-50でのEdge TPUのLatencyの 53 ms に対して、Efficentnet-Edge TPU-S では 5ms と10倍速いと。。
そして、精度もいいと。。。
そして、精度もいいと。。。
Efficentnet-Edge TPU-Lだと、Latencyは 25 ms 程度ですが、
精度は80％超えて、Inception-v4やInception-Resnet-2よりいいです。
精度は80％超えて、Inception-v4やInception-Resnet-2よりいいです。
Cloud TPU上で学習して、Post-training quantizationにて量子化して、Edge TPUで実行できるというのもいいですね。
Post-trainingquantizationは、July 2019 Updatesで対応できたんですからね。
