BOUNTY: Fully client-side auto-nounify with face-api.js

[socathie]

Executive Summary

• Project Overview: The ultimate goal is to create an application that can automatically mask all the faces (front or side views) with specified image assets.
• Bounty Overview: As the first bounty in this project, the goal is to reproduce a fully client-side version of auto-nounify by replacing the Google API with face-api.js.

Project Details

• Scope of Work:

1. Study the existing auto-nounify repos (see Reference Material below)
2. Modify (or create a new one) the project such that it is a monorepo.
3. Replace the server side process with face-api.js (its ssd_mobilenetv1_model and face_landmark_68_model) working purely on the frontend.
4. Fine-tune the results with the face_landmark_68_model, which is expected to be differently indexed from the full Google Landmark Detection model.
5. Update README.

• Expected Outcomes:

1. A well-commented monorepo (this repo), especially highlighting how to replace image assets and how to fine-tune the asset position
2. A minimal UI (like auto-nounify.wtf) for the tool
3. The application should produce similar (or better) results as the sample images
4. An updated README file with testing and deployment instructions

• Technical Requirements: TypeScript, any frontend framework you prefer

Qualifications

• Skills Required: frontend/full-stack development
• Preferred Qualifications: previous experience with facial recognition tools is a plus

Administrative Details

• Bounty Liaison: Cathie So (@socathie)
• Estimated Project Duration: One month after bounty commencement
• Project Complexity: Medium

Additional Information

• Reference Material:

1. face-api.js usage
2. auto-nounify app
3. auto-nounify backend
4. auto-nounfiy frontend

Submission Details

• Submission Instructions:

1. Claim this bounty by commenting on this issue (explain why we should pick you, including your professional experiences, or your involvement in our community)
2. Once the issue has been assigned to you, you can start working on the bounty
3. Submit the bounty by forking and making a pull request to this repo
4. Improve based on the evaluator's comments (if any).
5. The bounty is considered complete when the PR is merged.

Remuneration

500 USDT (Polygon)

[jwklo]

This is Jonathan Lo, I would like to join this bounty. I am a full stack web developer and mainly do chatbot machine learning tasks this few years. And I had also attended the events of 852dev.

[Lokman928]

Hi, I am Thomas, I am interest join this bounty. I am a fullstack developer and interest in facial recognition. In the past, I have join different event for 852dev.

[arku3]

Saw this on twitter today and I have code a quick POC using NextJS at https://github.com/arku3/auto-852devify. Would be very happy if it will help anyone to get started.

Edit:
Deployed to vercel for demo https://arku3-auto-852devify.vercel.app/

[socathie]

Hi @jwklo, the bounty is now assigned to you. Please observe the deadline of one month and deliver by Oct 30 EOD (HKT).

Since @arku3 was kind enough to provide a POC, we expect most of the time spent on this bounty to be dedicated to fine-tuning the algorithm such that it yields excellent results on the sample images.

[socathie]

Bounty completed with #4

Payment: https://polygonscan.com/tx/0x1fce1ded701693817a8a6883c3f88398cc30f8839e44143761b942a54e8c85ff