While a pre-trained open source int8 mobilenet_v2 is currently not downloadable, the team is working on making an int8 QAT model available. If you could check back again in a couple of months, that would be great.
In the meantime, here are a couple of options:
- Attached is a Python script that can be used to generate an int8 model with post-training quantization. Please note that the script currently uses random data for the representative data set.
- If you are ok with using a newer image classification model, then there is an int8 quantized EfficientNet-Lite at this location.