Develop AI-Market APP Proposal

[steven-jun-liu]

AI-Market APP Specification

Goals:

• Create interesting scenarios for users to interact with the public AI for fun.
• Prototype of AI Store. The AI Store idea is similar to APP Store, in AI-Market APP users can interact with the public AI and the open AI app.

Scenarios:

Scenario 1. Image

• Step1: User takes a photo,

• Step2: Call open AI to detect the image category. Face OR common Picture.

• Step3 A: If it is face, show public face AI list,

  • Face Attributes detect ( Baidu, Google..): Age, Color, gender, emotion, glasses..
  • Face generation like Google dream.
  • Other Face AI...

• Step4: User chooses the Baidu AI.

• Step5: Show the AI results and show a button “Try other AI?”. User can switch to other providers’ AI for compare.

• Step3 B: if it is common picture. Potential AI List:

  Brand logo recognition, Animal recognition, Plant recognition, OCR…

Architecture

• Client Lib
  • Billing Uses ATN token to pay services. For better performance, Raiden Network can be bridged the public blockchian.
  • Call AI The data flow of calling AI shoule be minizied the hop, at most transferred by DBot Worker once.
• APP Server A typical REST web services.

Develop Tasks

Orangize AI Category

For example,

• Face

  • Recogntion

    • Baidu, Google, ...

  • Generation

    • Google Dream

• Image

  • Brand Recogntion
  • Animal Recognition

Client Lib

Node/Python/Java Version Lib

APP Server

Reuse the architecture of AI Market Server

APP

First, deverlop iOS APP

[17Damon]

• react-native: https://facebook.github.io/react-native/
• Alibaba Ant Design Mobile: https://mobile.ant.design/index-cn
• I suggest using AI Market Server together

[pplam]

There are some other repositories:

• https://github.com/jondot/awesome-react-native
• https://medium.com/@ste.grider/component-kits-for-react-native-84eff4b321b9
• https://nativebase.io/
• https://github.com/react-native-material-design/react-native-material-design

[chrouzz]

I will go through all the different steps and define the underlying requirements.

Step1: User takes a photo,

• no requirement

Step2: Call open AI to detect the image category. Face OR common Picture.

1. Before calling an AI service to detect image category, a user need to have a user interface to select the AI service which can detect the image category. It can be a website which display all AI services.
2. To enhance user experience and to help him to find appropriate AI services, we can add tags to AI services, such as "image", "video", "audio" ect... If the AI users set the image filter we will only display AI which apply on image file.

Step3 A: If it is face, show public face AI list,

Face Attributes detect ( Baidu, Google..): Age, Color, gender, emotion, glasses.. Face generation like Google dream. Other Face AI...

If after using the detect image category AI we want to automatically suggest other API we need to have a "recommendation mechanism". To make this recommendation mechanism we need to do a data analysis on the result.

Step4: User chooses the Baidu AI.

Step5: Show the AI results and show a button “Try other AI?”. User can switch to other providers’ AI for compare.

Step3 B: if it is common picture. Potential AI List: Brand logo recognition, Animal recognition, Plant recognition, OCR…

edit: after talking to Alex I updated my comments.