Darknet for MobileNet v2

Introduction

With the help of MobileNet-Caffe and pytorch-caffe-darknet-convert, the MobileNet v2 pretrained model is transferred to darknet

Only the src/image.c and examples/classifier.c are modified, search for "mobilenet" to see what are changed

Evaluation

If you want to evaluate with single image, use

./darknet classifier predict mobilenet/imagenet1k.data mobilenet/test.cfg mobilenet/test.weights mobilenet/cat.jpg

Expected outputs (GPU=1, OPENCV=0):

19.65%: Egyptian cat
16.53%: tiger cat
10.64%: tabby
 7.38%: red fox
 6.09%: kit fox

If you want to validate with ILSVRC2012, please refer to pjreddie.com to make a valid list

The result is:

Note

• With the parameter of predict, the outputs will be different according to options combination in Makefile

• The network structure is not exactly the same to the paper, therefore the Ops is larger