Our Cloud model was based on ResNet and SVM based on standard frameworks. With project Pluto, the cloud model has been changed to support Mobilenet in line with edge computing needs. We need to make sure the model architecture for Cloud should not be chosen around a mobile architecture (i.e Accuracy Vs Latency) and believe we should keep the architecture optimized for each deployment options.
The VIsual Recognition team has been made aware of this per Devin's conversation with Matt Hill and Allie.
Why is it useful?
|Who would benefit from this IDEA?|
How should it work?