The latetest Recover V.2 model makes wonders to my old black and white photos.
But it tends to give the persons beard even if they dont have it in the picture
Attaches before and after of my grandfather and his friend after a badger hunt.
Hi.
Recover is the Ideal model for images such as yours however, there are a couple of caveats to bear in mind first of all, It is recommended to use images that are 1000px and smaller to get the most optimized result. Basically it is recommended that the image is one megapixel or less.
Therefore, if you use the Recover Model on larger images you have to expect that there maybe some undesirable results. Having said that, Recover has done a pretty good job with your images which can be easily rectified in Photoshop or another image editor.
But, you don’t want to do that, you want to fix it in Gigapixel so here’s a couple of tips
You know Gigapixel is great for upscaling image however, are you aware you can downscale aswell, then apply the Recover Model before upscaling again.
Simply open Gigapixel as a standalone application because, I’ve found it works better this way then load your image once your image is open within the Preview area to your right you’ll see the side panel and just below the navigation window you’ll see the Scale Options 2x, 4x, 6x and Custom, under that is the Scale Factor with a number on the right
Click on the Scale Factor to highlight it and type in 0.50 then, press Enter to make your image one half it’s original size or type 0.25 and press Enter to make your image one quarter it’s original size.
Now that your image is smaller, you can now apply the Recover Model and hopefully achieve the result you desire and finally Upscale your image once again by simply typing the number 1 into the Scale Factor or click on 1x option to increase your image back to its original size
Here’s a link from Topaz about optimising the Generative Models
Hope this helps
Hi.
To be honest, although I can’t see the original image in full it certainly looks as though the original image was incredibly blurry so, the Machine Learning interpolation can be forgiven for thinking the shaded area around the faces was let’s say a five o’clock shadow
Having said that, you must admit Face Recovery has done an incredible recreation that can certainly be fixed in Photoshop. After all that’s said and done any application that gets me 70, 80 or 90% to where I want to be is certainly a welcome asset
Besides, I know you have presented this image as an example however, have you considered using Redefine with a Creativity setting of 2 on this image to achieve a global enhancement then apply Face Recovery
I think you’re be pleasantly surprised by the results
Alternatively if you have access to Topaz Photo AI try using Super Focus in combination with Recovery Faces and the Sharpening Enhancement
Again I think you’re be pleased with the results
Don´t misunderstand me. I´m very pleased with the result and I understand it can be hard for the AI model to distinguish unsharpness from facial hair. I thought about it this way to help the AI model: Detect faces in the picture and let the user check/uncheck a box for each detected face if it is facial hair och blur. But I will try with smaller scale factors since my original scanned picture is large.
1 Like
