[Explanation] The person in the picture is named Sheng Jianzhong, a doctoral student at Huazhong University of Science and Technology and City University of Hong Kong.
He led a team of more than 10 members to design a set of artificial intelligence algorithms with fast repair speed and accurate restoration of details.
Just import the original image into the system, and the repaired image can be automatically generated in less than 1 minute.
[Concurrent] PhD student Sheng Jianzhong
That is, we get the photo. I see if your photo is very standardized. Is the face that needs to be repaired in the center of the image? If not, we will cut (take) it, cut it (image) to the center, and cut the target area. Come out, we save it after cutting it out, and put it directly into our program to run. The time is actually very fast, it may be less than 1 minute, and it will come out in about half a minute.
[Explanation] The system is called "AI Baby: High-resolution Image Restoration System for Family Search". Since March 2020, it has repaired blurry old photos for more than 1,000 lost children for free.
[Concurrent] PhD student Sheng Jianzhong
We think this matter is very meaningful. Compared with those fees, it is a petty profit. I think we, as a student team, do not need to care about this. , people who contact us by email, email us directly, we will repair them for free.
[Explanation] Sheng Jianzhong said that most of the old photos on the Internet are restored by manual PS.
The advantage of artificial intelligence algorithms is not only the fast repair speed, but also the more realistic restoration of details.
[Concurrent] PhD student Sheng Jianzhong
The first one we repaired is more clear than others, because our pixels are higher, the second one is for children, we have our own database, we have our own face database, so the accuracy of our repair is better Gao, the third one is that we target special application scenarios, that is, they need to print it on the flag, on the body, and we do a super-resolution reconstruction of the image, so the photos we get may be 10KB20KB , but we give them, we can do 150 megabytes, and even some we can do more than 800 megabytes.
[Explanation] Sheng Jianzhong said that the application of image restoration technology to family search has solved the core problems of blurred photos and low resolution in family search.
The clear photos made it possible to widely disseminate family tracing information, successfully turning isolated family tracing into social tracing.
[Concurrent] PhD student Sheng Jianzhong
Now we can fix it clearly, and after fixing it, he can print it on that kind of product, such as the love tape we make, such as the food packaging bags of some companies. After printing the product, it can be widely disseminated. , when it became wider, he mobilized the whole society to help his small family to find relatives, so that his family-style relatives search became a social relatives search.
[Explanation] At present, Sheng Jianzhong's team has published 2 papers at the top international artificial intelligence conferences AAAI and CVPR, and has won the 2021 China International "Internet +" Innovation and Entrepreneurship Competition Hubei Province Bronze Award, Hubei Province College Students Science and Technology Entrepreneurship Special, National Technology-based small and medium-sized enterprises have more than 20 intellectual property rights such as invention patents.
(Reporter Zou Hao from Wuhan, Hubei)
Responsible editor: [Li Yuxin]