v1.3
This beta version powered by Collective Knowledge can:
- Support basic DNN benchmarking (init+recognition) for Android 5+ (time and accuracy)
- Support for DeepScale Squeezenet 1.0 and 1.1, BVLC CaffeNet and AlexNet, and BVLC GoogleNet
- Share results via http://cknowledge.org/repo (see "crowd-benchmark Caffe library using mobile devices" scenario)
- Share mispredictions (images and correct labels)
NOTE: Don't forget to press 'Update' to get latest crowd-scenarios!