How does AI magic make old Chinese movies look new?

0


02:20

AI technology is giving old Chinese films a whole new look and giving restaurant owners a helping hand to navigate an otherwise slow and tedious process.

Before the age of digitization, film preservation was a mammoth undertaking involving the cleaning, repairing and patching up of the best surviving pieces of a film from bygone eras. The technicians had to manually transform the footage recorded on film into digital format, frame by frame. It was time consuming and laborious.

A frame is one of many individual images that, when put in sequence, creates a moving image.

“Traditionally, a technician could spend a whole day repairing only about 200 frames. A second of footage contains 25 to 30 frames. So you can imagine how long it would take to repair a film,” said Zhang Xiaoyun, associate professor at Shanghai Jiaotong University Medianet Cooperative Innovation Center.

Comparison between the old film (L) and the repaired film (R). / Shanghai Jiaotong University

Comparison between the old film (L) and the repaired film (R). / Shanghai Jiaotong University

But AI helps streamline and optimize the task. Algorithms can be trained to comb through massive amounts of material, analyze it, and extrapolate their knowledge to overlay old films with missing data and other inputs.

“The AI ​​restoration enhancement technology that we are developing is completely data driven. By learning from tons of material, the model knows what old films and repaired films look like. It can automatically repair. movies through its algorithm, ”Zhang told CGTN.

“It’s much more effective than traditional methods,” she added.

AI can reduce auditory and visual noise, emphasize colors, insert additional frames for smoother animation, and improve resolution.

In 2006, China launched a project to restore 40,000 old films shot on celluloid tape. Several film institutions and streaming websites, such as iQiyi, Youku, as well as China Film Group Corporation, have each developed image processing systems powered by AI.

The Medianet Cooperative Innovation Center of Shanghai Jiaotong University has cooperated with China Media Group (CMG) in recent years to repair classic movies and songs, most of which now meet 4K or 8K ultra high definition broadcast requirements. .

Comparison between the old film (L) and the repaired film (R). / Shanghai Jiaotong University

Comparison between the old film (L) and the repaired film (R). / Shanghai Jiaotong University

Archived films can suffer damage of various kinds. Scratches are common, and the plastic strips on which moving images are captured tend to lose quality and even warp over time. All this makes the restoration work more difficult because the results largely depend on the condition of the materials available.

The better and more detailed the images, the more likely the success of the AI ​​restoration is, Zhang said. But if the film is badly damaged, repairing it is almost an impossible task.

Film restoration is not only about technology, but also art and history. Technicians cannot rely on machines alone to recover conventional films, especially when it comes to colorization.

How the AI ​​transforms a black and white film into a technicolor version or brightens up the colors of an aged film is pure guesswork, based on its knowledge of real world objects. Sometimes the algorithm is successful. At other times, he misses the mark.

Zhang expressed the hope that artists and directors can get involved in the restoration work.

“They can provide us with colorization requirements for a keyframe, for example. Then we’ll know how to fix the rest,” she explained.

“AI is not omnipotent. We hope that in the future, AI will do 99% of the work, while humans will only need to cover the remaining 1%,” said Zhang.

National GPU platforms are urgently needed

Comparison between the old film (L) and the repaired film (R). / Shanghai Jiaotong University

Comparison between the old film (L) and the repaired film (R). / Shanghai Jiaotong University

Restoring old movies to AI requires a special type of chip, called graphics processing units (GPUs), for rendering video and editing images. These cards were originally used in video games to enhance their visuals, but are now used in a variety of applications, including mining cryptocurrency and restoring AI movies.

Currently, Chinese film restorers import most GPUs, with US company Nvidia dominating the market.

The Cooperative Medianet Innovation Center currently has around 200 students whose main job is to test and develop new AI models, including preprocessing, decoding, enhancing and outputting encoding sequences. “We have around 100 GPU training cards – most of them imported. It cannot meet the large practice and research work of the whole center,” Zhang said.

Huawei offers technical support to the Cooperative Medianet Innovation Center at Shanghai Jiaotong University. / CGTN

Huawei offers technical support to the Cooperative Medianet Innovation Center at Shanghai Jiaotong University. / CGTN

The center is now working with Chinese tech company Huawei, which provides them with its Shengteng GPU platform and a server that can support eight GPU cards.

“Through this cooperation, Huawei and we hope to develop our domestic GPU platform and make full use of our GPU resources,” Zhang said.

(Editing by Nadim Diab, Video by Zeng Hongen)



Source link

Share.

About Author

Kehoe Young

Comments are closed.