mosAIc

Face Publishing Approval Setting
Take a photo on the app and share on SNS!
Posted on SNS. The pic has been processed by AI mosaic engine with reflecting user's will on Approval Setting.
Check your face and Decide whether your face is gonna be mosaicked or not.
The result after changing user status!

Inspiration

Have you ever been shared a photo you are in without your permission?

Many people do not feel comfortable to be shared of the pics online without permission. We call this uncomfortable experience “Photo Harassment” and this is one of the biggest social issues these days in Japan. To solve the problem, we created this app named “mosAIc.”

What it does

mosAIc! will provide automatically creating the mosaic to mask people’s faces which are not permitted.

All user have to do is just 3 things.

Step1. Setting.

There are just 2 setting for using the app. The one is to register your front face view photo. The second one is to configure whether it’s ok to show your face or not when someone post a pic you are in. mosAIc! have 3 options for this: Always OK, Always NO, Ask me every time.

Step2. Take pics on mosAIc!

Step3. Share them on SNS!!

With just one picture which you set on user setting, the AI system we created will recognize who they are in the picture.

How I built it

User Interface

We created a Simple SNS application for showing the pics which are processed by Original Image Processing Engine named by “mosAIc engine”.

This interface has 3 functions which are camera function, face publishing approval setting function and SNS platform, so you can take photos and share them on the same app.

This application will be provided as web application.

Back-end

We use RDB and Object storage in cloud service for building this application.

RDB holds user’s data such as name, face image file path data on the Object Storage, approval status. Using each data, the app will operate mosaic processing.

Object storage holds image data which user posted and Json file which has face coordinate information.Back-end server works to call API engine we created and a cloud service whenever users do anything.

In the case that a user, who set on ”Ask me every time mode” which is one of the face publishing approval setting, decided to show the face, then the system will remove the mosaic.

Original Image Processing Engine

We created “MosAIc! Engine” which consists of two AI modules such as Face Detection Module and Face Classification Module, and consists of one image processing module Mosaic Processing Module.

With just one picture which you set on user setting, the AI system we created will recognize who they are in the picture.

Challenges I ran into

To create the application’s UI, we decided to use Vue.js. Some of our team had no experience about how to use the language, but we challenged and eventually we mastered how to develop with the language through Junction Tokyo.

And also, this is our first time to use Object Storage API. We used Cloud Object Storage provided by IBM Cloud. It was a little bit tricky to use because we had never used it. So it took a lot of time to use it but we made it!!!

Accomplishments that I'm proud of

We are proud of two stuff. First, the app implements the idea which the app masks or removes mosaic in the picture people uploaded on SNS depends on people’s will.

Second, we are proud of the accuracy of creating Original AI image processing engine named ”mosAIc engine” which implements three parts which are face detection function, face recognition with just only one picture, and mosaic processing to mask people’s faces.

What I learned

About AI..

We learned how to calculate similarity of face with just one picture which user set. To speed up the response time for image processing, we made each processes changed from centralized process to separated process.

What's next for mosAIc

To use the original mosaic engine we created, we would like to apply it to video as our future prospect. We tried to make it but we did not have enough time this time. Therefore, we will make it next time!

In addition, we would like to implement sentiment analysis AI on the mosAIc engine and would like to process masking to stamp people’s face with the AI because this would be able to express people’s feeling without showing real face instead of mosaic.