Pure Technologies <> Utilizing Computer Vision + Machine Learning for Natural & Artificial Toning

@Mr_Spud said:

@treybenedict said:
Also side note - I’ll be at Long Beach Thursday & Friday to talk about the model, theory and any other details that anyone wants to ask.

Also will have goodies for anyone who finds me

Are you going to have a booth?

No, I'll be walking the floor likely. Thursday-Friday. I'll have a booth at Summer FUN & ANA.

@split37 said:

Can you run these through your program? Peace dollars tone much less intensely than Morgans. I believe having them compared on the same scale may be skewed. It may serve well to have a rating system per the series and not compared with ALL series.

Yes, I will run everyone's after the app is released. I have been super busy with development and fixing bugs

@silviosi said:

@ treybenedict

Very interesting study project. I put your study from the point of view of humans. You use in your study RGB spectrum mainly for silver toning. This spectrum is base on saturate colors and the humans perceive non-saturated color on the scale of 390 to 750 nm and can not see infrared and UV spectrum.

So due to above considerent how your algorithm will can calculate the eye appeal? Then the toning is direct proportional with the environment (in this case the Ag 900) so how the soft will decide if was naturally or manufactured? Do you guys there can resume to analyzes of 390 to 750 nm unsaturated colors?

Maybe you will find hard my questions, I know. But as a person with many academic studies I put questions which can open others doors to the perfection.

All my best: Silvio

If I understand this correctly, I based the analysis off saturated visible light colors. Infrared & UV are likely not detectable given the image quality and conditions available (since the photos were meant to capture eye appeal).

I can breakdown RGB into wavelength patterns, which I think is very interesting as well. Thankfully cv2 is powerful enough to do this with just enough memory on our upgraded servers.

The basis of NT/AT/QC is on previous training data that was gathered, and then predicted for the new image using a CNN.

@silviosi said:
As I say: This project it is very interesting. Breakdown into waves pattern I think will help you at the best. 20 years ago I was mad again the info guys because I can not see the spectrum on radio-oncology cellular radio. In time was not this day technology. I think CNN is the best for the moment and I hope better will come and you will adjust.

Me I find it is a great, brilliant concept.

Thank you sir Appreciate your kind words.

@bretsan said:
Wow, your app is awesome. Thanks for analyzing my coins. I’m going to try out some others.

Thanks! Happy to share with you.