/caffe-java

Using java-cpp to build a caffe based java classifier.

Primary LanguageJavaMIT LicenseMIT

Description

This is an image classification tool which recognizes models in images. Yes, it is that cheesy.

Building

Building with docker (results in 7gb image):

  • Install & Run docker.
  • Install maven.
  • Run mvn install in main directory.

Building without docker:

  • Install maven.
  • Install dependencies as suggested in Dockerfile
  • Run mvn install -DskipDockerBuild in main directory.

Running with docker

Running with docker:

  • Start docker
  • Start container with sudo docker run -it -v ${someDirectoryOfImages}:/images ae/caffe-java /bin/bash

Running in general

Use java -jar caffejava.jar [flag] [file]. where flag is either '-i' or '-m', with '-i; indicating you are classifying a single image, of '-m' indicating you are want to process a movie file. If you classify an image file the output will be writen directly to the screen. If you classify a movie the file will be split into different frames which will be classified and writen to a tempdir in the /tmp (or in the windows equivalent) directory.