inceptionServer

A web server to assign labels to images using tensorflow inception model.

It will show a random image, and its labels. The image set is loaded from local filesystem --imgdir. New images can be added to this directory, and will be shown in the page.

Run it

Run it based on docker image

docker run -p 8080:9527 -d beekman9527/inceptionserver

Then It can be acccessed through http://localhost:8080/ .

Or load your own set of images for the server via:

myimgdir=/my/img/dir
docker run -p 8080:9527 -v $myimgdir:/tmp/imgs/ -d beekman9527/inceptionserver

Images can be found from ImageNet.

Build it

Pre Requirements

• Golang

• Glide (package management tool)

• tensorflow for Golang

  According to the instructions of Installing tensorflow for Go, this script will do the job:

Build local runnable binary

With the following commands, a executable binary will be built in inceptionServer/_output/inceptions.

cd inceptionServer
sh scripts/init.glide.sh
make build
sh scripts/run.sh

Then run this binary with following commands:

_output/inceptions --v=2 --logtostderr --modeldir=./model-data/inception/ --imgdir=./imgs/

A web server will be listening on port 9527. Access it via http://localhost:9527.

Build a container image

sh scripts/build_img.sh

or build a container image contains only tensorflow runtime lib and the binary file of this server

sh scripts/build_img_lit.sh