Caffe2Go evaluate caffemodel with Golang
Command line interface
./caffe2go -i images/plane.jpg -m models/nin\_imagenet.caffemodel -l labels/synset\_words.txt -s 224 -mf means.txt
Options
age of ./caffe2go:
-cpuProf string
Filename for CPU profiling.
-i string
Path for image.
-l string
Path for labels.
-m string
Path for caffemodel.
-memProf string
Filename for Memory profiling.
-mf string
Meanfile path
-s uint
Input Shape
Use the library on your own software
package main
import (
"fmt"
_ "image/jpeg"
_ "image/png"
"github.com/Rompei/caffe2go/c2g"
)
func main() {
caffe2go, err := c2g.NewCaffe2Go("lenet.caffemodel")
if err != nil {
panic(err)
}
output, err := caffe2go.Predict("mnist_zero.png", 28, nil)
if err != nil {
panic(err)
}
for i := range output {
fmt.Printf("%d: %f\n", i, output[i][0][0])
}
}
Now supports the layers below
Convolution
Pooling
ReLU
FullyConnected
Dropout
Softmax
LRN