Currently this is only tested on Mac platform. This experimental project is based on the following projects.

1. https://github.com/cmusatyalab/openface, Apache 2.0 License. Modify its Dockerfile, https://github.com/grimmer0125/openface and used its trained face NN data and python library in this project.
2. https://github.com/yyosifov/image-viewer, ISC License. Most of UI and Electron related codes of this project are modified from it.

This electron-face-finder application will automatically launch the docker container of openface which has a python websocket server running.

Known Issues:

1. It seems that sometimes electron side will send partial data to the python sever running in the previous docker container. It may happen when you close the app and immediately restart it and operate it very soon.
2. Slow python WebSocket in Docker. Need to debug it.
3. • cpu usage is too high. (maybe the code in openface can be as a reference)
4. • memory usage is a little high
5. sometimes html image object will load an image file in a wrong way. Such as rotating a 90/180 degree somehow. It is exif issue.
6. If the folder contain too many images(may hundred of files), "out of memory" will happen and throw exception, net::ERR_INSUFFICIENT_RESOURCES.
7. Need to hanle the exception due to the symbolic link of a file.

Todo list

1. add source/targetImage info in the ack packet from server. It is to handle interrupt case. E.g. Change target/source when the process is not finished.
2. stop button.
3. study how to debug main process when this app is as a package app.
4. Add connecting indicator to block UI operation and progress bar.
5. Clean the sending queue when reseting
6. Add Mac app signing.
7. Support multiple faces in an image file.
8. try to support the image path having "%"
9. use readAsArrayBuffer to read a image file and send to the server instead of the current way, which adapts FILE -decoding> Image object -draw/copy> canvas -> jpeg-base64-DataURL then send. memory/cpu related.
10. Linux, Windows support
11. When getting the response from the server, use dictionary to lookup the image object instead of using array. cpu related
12. Python 2.7 -> Python 3.5

Before run in in development environment or download a packaged app, you need Docker for Mac.

Type sh checkinstallDocker.sh to install it. Or you can download it from https://www.docker.com/products/docker#/mac and then install it.

How to Run in development environment

Install additional dependencies before running locally

After installing it, type docker pull grimmer0125/openface-nostartdemo to install the needed Docker image. Or you can use Docker-UI-tool kinematic to search the above docker image (openface-nostartdemo) and install.

then type npm install to install NPM packages.

Run method 1

1. type npm start in terminal. But this way does not have any terminal logs of server side, you need to use kinematic to monitor the server logs from Docker. Or use shell script to attach/exec into the Docker.

Run method 2- Run with monitor server logs directly

1. type make run in terminal. In this way, you can see server logs. But you can not use ctrl+c to close Electron-UI app. You need to close it inside Electron app.

Run Production packaged app:

Install additional dependencies before running locally

After installing it, type docker pull grimmer0125/electricface(web:https://hub.docker.com/r/grimmer0125/electricface/) to install the needed Docker image. Or you can use Docker-UI-tool kinematic to search the docker image (electricface) and install.

Download the distributed binary packaged Electron app, FaceFinder.dmg form here.

https://goo.gl/5IzCUj

When use this applicaiton, please do not quit Docker.

Face Finder/Matcher

A cross-platform Face Matcher based on Electron. It runs on OS X, Windows and Linux. It provides an easy interface to browse a directory of images using the left and right keyboard buttons.

Caution

1. It will search all the sub-folders and consume a lot time (1 file-1MB consumes 5s).

Preview

• OS X: (dog picture is from https://github.com/yyosifov/image-viewer)

Features

• Find the matched face images. From the menu, file->open a source file, select a source image including 1 face. Then click 'open' of menu or click open button , to select folder/any other file. Later it will show matched images having the same face.
• Open a Directory and browse it's images
• Open an Image File and browse the images inside the same directory
• Fast navigation through the images via Left/Right keyboard buttons
• Rotate an image clockwise/anti-clockwise
• Save/Move the image to a new location