I am not an expert either, so I can not comment if the loss of signal is excessive or not. However, since I am not proficient with DBE, the amount of time I would save is much more than a half hour. Thank-you for sharing and educating other hobbyist like myself.
Just be wary the AI version of Grax can erode the good signal significantly on certain images… It’s pretty hit or miss. I put a video out where I use it in my workflow. But on some targets, especially more faint gas structures it can completely destroy the image. When it works it is awesome though for sure
It's debatable, I think I can trade a portion of my signal for a perfect flatness. If you find an image on Astrobin with a lot of signal but bad residual gradients that weren't removed by classic DBE, you'd automatically label the image as not good, no matter the signal in it. We can shoot more integration time to cover the expenses of GraX but I think it's worth it.
It’s definitely fun to use and super quick. Though I can’t imagine the professionals using graxpert and sacrificing signal. These days I always clone the image and compare my effort with DBE to graXpert. Most of the time I prefer DBE because it’s a different kind of fun 😊
I have had posterization using GrXPert on occasion. I have tried several work-arounds but no luck. Reverted back to DBE or ABE. The image seemed to improve with HT [not using GHS yet] but something seems to happen maybe with NoiseX and GraxPert? Whatever, user beware.
I gave this a try last night and I'm not a fan. First image was SHO composite of NGC 1499 (no processing, just combination of 15hrs equal SHO mono) which turned the image into garbage. The developer confirmed on Astrobin that the AI model is not trained for SHO images, so you should use on mono images first before combination. I tried it on my OIII data from NGC 1499 which has no well defined structure, and the AI just took out all the OIII signal because it thought it was a gradient. So far, no good. My second image was a mono RGB composite of M45 with a strong gradient (no other processing performed). The AI did not work so well and left a red gradient throughout the image, whether I used subtraction or division. So for these images, GraXpert did not work for me. Whilst DBE takes some time, about 15 minutes per image, I can usually eliminate about 99% of a gradient from my image so I will continue using DBE until GraXpert has been significantly improved (or perhaps Russell beats them to it with a new GXT tool 😉).
I always love to test new tool as they come but in this case I reverted to DBE to! RC Astro has a GradientXterminator (not sure about the name!) tool that works in Photoshop. I used it prior to PixInsight and it was great 😊
@@SPACEJUNKastro Ah yeah, I forgot about that. I don't use photoshop (I use Capture One for regular photography) so I've never had the chance to try it out!
DBE for life!
I am not an expert either, so I can not comment if the loss of signal is excessive or not. However, since I am not proficient with DBE, the amount of time I would save is much more than a half hour. Thank-you for sharing and educating other hobbyist like myself.
Just be wary the AI version of Grax can erode the good signal significantly on certain images… It’s pretty hit or miss. I put a video out where I use it in my workflow. But on some targets, especially more faint gas structures it can completely destroy the image.
When it works it is awesome though for sure
Exactly. Use with caution!
It's debatable, I think I can trade a portion of my signal for a perfect flatness. If you find an image on Astrobin with a lot of signal but bad residual gradients that weren't removed by classic DBE, you'd automatically label the image as not good, no matter the signal in it. We can shoot more integration time to cover the expenses of GraX but I think it's worth it.
It’s definitely fun to use and super quick. Though I can’t imagine the professionals using graxpert and sacrificing signal. These days I always clone the image and compare my effort with DBE to graXpert. Most of the time I prefer DBE because it’s a different kind of fun 😊
I have had posterization using GrXPert on occasion. I have tried several work-arounds but no luck. Reverted back to DBE or ABE. The image seemed to improve with HT [not using GHS yet] but something seems to happen maybe with NoiseX and GraxPert? Whatever, user beware.
I gave this a try last night and I'm not a fan. First image was SHO composite of NGC 1499 (no processing, just combination of 15hrs equal SHO mono) which turned the image into garbage. The developer confirmed on Astrobin that the AI model is not trained for SHO images, so you should use on mono images first before combination. I tried it on my OIII data from NGC 1499 which has no well defined structure, and the AI just took out all the OIII signal because it thought it was a gradient. So far, no good. My second image was a mono RGB composite of M45 with a strong gradient (no other processing performed). The AI did not work so well and left a red gradient throughout the image, whether I used subtraction or division. So for these images, GraXpert did not work for me. Whilst DBE takes some time, about 15 minutes per image, I can usually eliminate about 99% of a gradient from my image so I will continue using DBE until GraXpert has been significantly improved (or perhaps Russell beats them to it with a new GXT tool 😉).
I always love to test new tool as they come but in this case I reverted to DBE to! RC Astro has a GradientXterminator (not sure about the name!) tool that works in Photoshop. I used it prior to PixInsight and it was great 😊
@@SPACEJUNKastro Ah yeah, I forgot about that. I don't use photoshop (I use Capture One for regular photography) so I've never had the chance to try it out!