AI ID photo tool is on GitHub hot list, with built-in postgraduate and civil service examination visa templates generated in one click

Yishui from Aofei Temple
Quantum Bit | Public Account QbitAI

Recently, this AI ID photo tool has become very popular~

Ben casually uploaded a photo of Harry Potter, and suddenly, a blue-background ID photo was generated!

More importantly, this tool is currently free and open source , and everyone can use it immediately.

The tool is called HivisionIDPhotos . It has been on the GitHub hot list in recent days and has currently received 3.2K stars.

In addition to changing the background, it also supports lightweight cutouts, selection of standard ID photos of different specifications and purposes, and six-inch layout photos for easy printing.

Netizens who have experienced it have expressed their feelings:

One of the examples of AI's practical skills. In the future, just take a photo and AI will help you make it.

Musk, the first batch of "experience users", has already been arranged.

Some people even joked online that they have recommended it to Lao Ma and will use it when he takes the civil service exam (doge).

By the way, the author of this artifact is a doctoral student from Xidian University.

Let me try this for you guys

Next, we will give it a try for everyone.

The project author has provided a web demo , which is currently available in Chinese and English. We will go directly to the Chinese version.

This first level, of course, tests the AI’s ability to cut out images .

The hypothetical scenario is "using daily photos to create an ID photo for a formal occasion" to see if this tool can meet unexpected needs.

Please note that this tool does not support formal attire yet , so we still have to bring formal attire when uploading photos.

For example, here’s a photo with a slightly cluttered background:

To be honest, judging from the result, the cutout effect is still quite obvious, and it makes people hesitate to use it in formal occasions.

But if we change to other photos, just judging from the cutouts, the effect seems to be pretty good?

Judging from the examples provided by the official, it seems that the original image itself is of very high quality.

So I realized that, next time we take an ID photo, we can use this tool to change the background and size, the main purpose is to save money.

At present, some common sizes and uses are arranged, such as civil service examination, teaching qualification examination, college students' level 4 and level 6 examination...

There are also six-inch layout photos for your convenience in printing.

BTW, this tool currently provides API services based on FastAPI and also supports Docker deployment , allowing users to quickly build and run services in different operating system environments.

The project can run on Linux, Windows and MacOS platforms, and supports Python 3.7 and above

According to the author ZeYiLin:

The cutout model is fine-tuned by MODNet + open source ID photo data on the Internet, and converted to ONNX to accelerate reasoning; the face detection model uses MTCNN, which is also converted to ONNX. The running time of the whole set of code on the CPU is acceptable (but there is still a lot of room for optimization).

Next, judging from the project homepage, the author also plans to provide beauty and dressing functions .

In short, judging from everyone's experience so far, it's still a little bit difficult to take a photo for your ID card. (You also need to rely on luck)

But it's a good idea and can be used by individuals in different occasions.

The author is from Xidian University

The project author , ZeYiLin , is a first-year doctoral student at Xidian University, under the tutelage of Professor Wu Jiaji of the School of Electronic Engineering .

He is also one of the founders of SwanLab and SwanHub open source communities.

According to his personal homepage, he is passionate about full-stack work and loves machine learning, UI/UX and top-level product design.

According to what he posted on Zhihu, the algorithm involved in HivisionIDPhotos was the result of his previous participation in the "Internet + Innovation and Entrepreneurship Competition".

This is the code used in a small program I developed for making ID photos. It has been used by nearly 1 million people in China.

He said frankly that the reason for choosing open source now is:

With the advent of the AIGC wave, the scene of ID photo production, which is very biased towards the implementation of image code engineering, is a bit old.
At that time, there was an idea to make an online Haima font.

Finally, I’ll give you a trial link at the end of the article. Welcome to the comment section for feedback~

Online demo experience:
https://swanhub.co/ZeYiLin/HivisionIDPhotos/demo

Open source code:
https://github.com/Zeyi-Lin/HivisionIDPhotos

Reference links:
[1] https://x.com/tuturetom/status/1830778702751506916
[2] https://x.com/imxiaohu/status/1829870064352575576

Click here ???? Follow me, remember to mark the star~