cool hit counter You like "HD No Code" now you can remove the mosaic with one click_Intefrankly

You like "HD No Code" now you can remove the mosaic with one click

Ever since video cameras were invented, mankind's quest for picture quality has never stopped, with resolutions ranging from, 320p to 480p, then 720p, 1080p, and now 2k, 4k.

(320p to 4k schematic)

In addition to the gradual increase in resolution, the video frame rate has also improved, from 25FPS to 60FPS , then to 240FPS and beyond.

Videos taken at high frame rates are particularly stunning when slowed down, such as the slow motion of bullets.

Like the moment the balloon explodes.

How high of a frame rate do you need to shoot the bullets? Take the most common bullet, at the speed of sound 340m/s, the shooting range is about 1 meter, it takes 0.0029 seconds for the bullet to complete this 1 meter, if you can shoot 1000 pictures per second, then you can shoot 2.9 pictures of the bullet, that is to say, if the frame rate is below 1000FPS, it is basically useless to shoot the bullet, you can't even find the ghost shadow of the bullet.

But with AI, maybe you'll be able to shoot cool slow-motion videos, too.

Recently, NVIDIA (the one that sells graphics cards) showed the public an amazing AI they developed with MIT and Aalto University in Finland, an AI that can take low frame rate video and turn it into high frame rate video out of thin air.

The following motion picture was taken with a 30FPS camera, and after slowing down 8x, it looks like this.

30FPS slow playback at 8x equals only 3.75 frames per second, stuck straight into a slideshow 。。。。

And look at the results after the Nvidia AI treatment: the

Doesn't it feel much smoother?

It's not raw, it's not laggy, and it's even optimized for clarity of detail on the screen.

In the following motion picture, the girl on the left is the original film, after 60FPS slow playback 8 times, the girl on the right is the processed effect, the comparison can be found, the left girl's hair and dress have a very obvious snag, the right processed girl is much smoother.

Here's a breakdown of what makes this technology from Nvidia great.

For an original 60 fps video to remain smooth after slowing down by 8x, a non-existent 7 fps image needs to be inserted between every two frames in the original video.

These 7 frames of images are pressed into existence, so how would they come about? Should we force Nvidia to pull out 7 frames of non-existent images for you? So, can you force a 7 frame image insertion by counting the pixel changes? Of course not, it would look raw and unnatural.

So, how did Nvidia solve it?

NVIDIA AI learns over 11,000 videos of everyday life and sports activities at 240 frames per second, and knows all kinds of motion trajectories, such as how your arms move after you fall, and how your legs move, all of which it knows, and can then brainstorm the missing frames and add them to you based on "experience".

Feel the video

The framerate is made up, but can the picture quality be "made up"?

Of course it can, and the AI is also able to easily remove the subtitles from the screen.

It can take the parts of the photo that are blocked and give you them back through an algorithm.

In the image below, the face of the character on the left is blocked, and after AI recognition senses that something is missing from the face, it uses logic to fill in the missing part.

It also makes it easy to turn low-resolution images into high-resolution.

We can also tune high resolution using ps, but the way ps increases the resolution is mechanical, and while it looks like the pixels are higher from the parameters, it's blurry.

The AI, on the other hand, is much smarter about what it does, and through massive image training, it knows what something looks like and can still be very clear, if not clearer, after bringing the image resolution up.

And the other big guy who's also working on this technology - Google - is even scarier.

Take a look at this picture below

Do you think this picture is still salvageable?

What a save, it's not just a mosaic!!!

Rangoose 。。。。

Google hardwired it back to look like this.

I'm so impressed!

As for your concern about "that kind of code" well, Nvidia has not yet fought it head-on, while Google Brain's AI has really tried to do so, Google can automatically turn a mosaic of only 8 * 8 pixels into a clear image, that is, through Google Brain's AI, you do not have to have no code in mind, you can also see HD.

This technology is indeed also a boon for those who retouch pictures in post until their hands are weak, however, there are concerns that the AI automatically removes the watermark, which could be harmful to the photographer's copyright.

1、Remember the takeaway wars from a month ago The latest survey is out who won in the American Hunger Drop
2、How to hunt the duel configuration requirements odd tour accelerator full range of configuration requirements detailed
3、Following in the footsteps of the past and the future cumGD Blueprints 2018 Annual Meeting was held
4、Whats the front end How do I get started with the front end
5、web front end JS check password plugin automatically detects password strength comes with source code

    已推荐到看一看 和朋友分享想法
    最多200字,当前共 发送