Step by step process to generate 3d Photography of any image with an AI (CVPR 2020)
Generate a 3d image with any of your pictures. Try, be amazed, be inspired.
I am an enthusiast in AI field. Just looking for intresting things and learning along the way. I was looking at Two Minutes Paper’s video and found this topic interesting to explore. I couldn’t resist trying it on my own image. So if you are here to convert your still image into 3D photo with this AI. I hope this article satisfies that hunger of yours.
At the final you will be able to:
- Create an awesome 3D image from your still image
First, Go to Runtime > Change Runtime Type > Select GPU
You can see the play button in Google Colab to run your code in the notebook. So everytime you see them it means you can run them pressing the button.
Now, click the first play button and you will see the code running and installing or if already satisfied it will say “Requirement already satisfied”
Now run the training code as well.
Run switch off off-screen rendering code and when you run the other one you will be asked to upload a picture. Please be advised that you’ll need to grant Google Drive access to the folder first as given in image below. Go ahead and choose an image that you’d like and press. After uploading you’ll see file uploaded as given below:
Note: You have to enable acess to Google Drive from files. Click drive icon at top (3rd icon) to see the permission and give access accordingly.
I suggest you to delete all other files from /images and /video folder before moving to Step 4.
This is the step where you execute the 3d photo inpainting.
Just run(press play) the remaining codes:
!python main.py — config argument.ymlfrom google.colab import drive
You’ll be asked for authorization like the following. Click the link that says “Go to thie URL in a browser”. And you’ll get a code. Copy code and paste it.
After the code runs successfully you will get final video in /video folder.
Something we never thought possible is turning out to be true. This shows a new opportunity in the imaging arena. We can be hopeful the future will be better. It's not 100% accurate but it's near there.
I hope you learned something good or get inspired by this. Thanks for reading.
Title: 3D Photography using Context-aware Layered Depth Inpainting
Author: Shih, Meng-Li and Su, Shih-Yang and Kopf, Johannes and Huang, Jia-Bin
Like What You Read?
If you want to read more informative articles like this one, my Email digest is just the thing for you.