Models accuracy on ImageNet 2012 val - BVLC/caffe GitHub Wiki
Simple comparison of Model Zoo models. Single center crop from 256x256 warped image, mean value = [104 117 123].
Name | Top-1 | Top5 |
---|---|---|
BVLC AlexNet | 0.569 | 0.801 |
BVLC CaffeNet | 0.571 | 0.801 |
NiN* | 0.567 | 0.795 |
VGG-DevilS | 0.612 | 0.834 |
VGG-19 | 0.685 | 0.885 |
VGG-16 4x | 0.706 | 0.899 |
BVLC GoogLeNet | 0.687 | 0.890 |
Princeton GoogLeNet | 0.672 | 0.881 |
- mavenlin`s Network in Network model is kindly converted to current caffe format by ronghanghu