Product Photography question

[Planetdune]

Hello.

I got thousands of product images. These are made using a DSLR + photobooth. The camera is calibrated to the white of the photobooth background and all images have a universal white behind them. I like using auto levels to enhance the details of the products but I notice this messes up the white backgrounds. Sometimes it is all okay but other times they become blue-ish, grey or even green tinted. I know I could select only the product, then apply the auto levels only to the product itself and thus not affecting the background white. The issue here is that with thousands of images to process and most having small or thin details, it would make the selection process take too much time.

My question, what is the best way to quickly add auto levels to an image without that affecting the white background?

[carl123]

6 hours ago, Planetdune said:

My question, what is the best way to quickly add auto levels to an image without that affecting the white background?

The 2.6 beta has a new option called Select > Select Subject which "should" select only your product image, you can then apply auto levels to that part of the image only

 

6 hours ago, Planetdune said:

I got thousands of product images

You can run a batch job incorporating the above commands recorded in a macro

[Planetdune]

Tried the Select Subject in the 2.6 beta. It does not work, unfortunately. The selections it makes are all over the place but never correct. It's strange because to the human eye, this looks very obvious, it is a product completely surrounded by a "as-close-to-white-as-I-can-get-it" background. But no, it selects logos on the product or only a part of the product etc.. I am close to giving up and accepting I will have to make do with blue and green and grey backgrounds here and there.

[R C-R]

2 hours ago, Planetdune said:

It's strange because to the human eye, this looks very obvious, it is a product completely surrounded by a "as-close-to-white-as-I-can-get-it" background.

What a human may see as an obvious object involves not just what the eye can see but also how they have learned to identify things through personal experience -- IOW, how the brain has been trained to interpret the data fed to it by the eye. (That's why visual stimulation is so important for a baby's early development.)

So for an app to come close to duplicating that requires an AI trained on a very large selection of images & considerable processing power. It isn't too surprising that the Affinity beta isn't as yet very good it that.