MoI discussion forum
MoI discussion forum

Full Version: Funny thing at the opposite of Moi ;)

Show messages:  1-4  …  405-424  425-444  445-464  465-484  485-504  505-524  525-544  …  3005-3010

From: Frenchy Pilou (PILOU)
21 Nov 2018   [#465]
Blue Morning :)


From: mkdm
21 Nov 2018   [#466] In reply to [#465]
Pilou you're getting better and better on this kind of digital art :)
From: Frenchy Pilou (PILOU)
21 Nov 2018   [#467]
Funny to explore so many possibilities! ;)


From: Frenchy Pilou (PILOU)
22 Nov 2018   [#468]
For test a big city format! :)






From: mkdm
22 Nov 2018   [#469] In reply to [#468]
Last image is super cool!

If you could do some of these stuff in print resolution (at least A3 300dpi 4961x3508) I think that you could do a possible business.

Think about it.

Clearly in that case, the images should be very cool and not something "already seen".

Ciao!
From: Frenchy Pilou (PILOU)
22 Nov 2018   [#470] In reply to [#469]
Maxi possible with Magica is for the moment 2048 *2048 ;)
From: mkdm
22 Nov 2018   [#471] In reply to [#470]
Check it out: https://topazlabs.com/ai-gigapixel/

It's incredibly powerful!
From: Frenchy Pilou (PILOU)
23 Nov 2018   [#472] In reply to [#471]
Yes another deep learning application! ;)

Here seems free to * 4 for 15 megas max image! ;)
https://letsenhance.io but only 5 images!

https://github.com/alexjc/neural-enhance free but lines' commands so not very friendly UI!

cool articles
https://engineering.flipboard.com/2015/05/scaling-convnets
https://medium.com/nanonets/how-to-use-deep-learning-when-you-have-limited-data-part-2-data-augmentation-c26971dc8ced
From: mkdm
23 Nov 2018   [#473] In reply to [#472]
My two cents about this "ai" topic.

I'm a software developer and terms like "artificial intelligence" "deep learning" "machine learning" and so on, are really, really abused.

Don't forget that all that words are simply what is always existed from the beginning of the computer era, that is: something, some software procedure, written by more or less valid or even genius programmer/engineers, and this software leverage on a certain kind of hardware.

Nothing more than that: are software, ok very complex software, written by people who leverage on scientific knowledge that is already spread into many other disciplines.

A "learning software" is nothing more than a very complex software that follow a predetermined path (complete path) to achieve a wanted result.

The grate power today is that now we can run these kind of software on very capable hardware, very powerful.

No magic behind, no special or "intelligent" things.
But always the same old story: people who try to solve real problems using a computer programming language.

When I see or use a very complex software that give me the "illusion" of some sort of "intelligence" my first tough always goes to the smart people who created it.
They are "smart" "intelligent" "learning".
They, not the machine, the piece of metal.

Let's take the example of Alexa or Siri or other smart "services".

When I say "Hey siri, bla bla bla..." the real truth is that Im' talking indirectly to all the people that engineered that comple software and I say them: "Hey people who are listening, con you help me?"

This is the real fact of the matter.

Great software, complex software on super powerful hardware, created by very smart and capable people who work and strive to create these kind of things.

Have a nice day.
From: Frenchy Pilou (PILOU)
23 Nov 2018   [#474]
with https://letsenhance.io/
So small JPG --> big PNG result --> save as JPG for more reasonnable size for load on a forum ;)
(730 KB (JPG) --> 114 Megas (PNG)--> 15 Megas (JPG)

2048 * 1706 --> 8192 * 6824

Second image same but starting with a PNG (so no lose information like JPG)
(6 Megas (PNG) --> 114 Megas (PNG)--> 15 Megas (JPG)



Second image


From: mkdm
23 Nov 2018   [#475] In reply to [#474]
Very very good result I see here...

So...you can now start your new business selling your Voxels renderdings? :)
From: Frenchy Pilou (PILOU)
23 Nov 2018   [#476] In reply to [#475]
No only for pleasure of the eyes! ;)

Curiously seems exactly the same result when starting with a JPG image or an PNG image!
2048 * 1706 --> 8192 * 6824

JPG was a transform of the PNG original before process!


From: mkdm
23 Nov 2018   [#477] In reply to [#476]
@You : "No only for pleasure of the eyes! ;)"

Pity :)

Ciao!
From: Frenchy Pilou (PILOU)
23 Nov 2018   [#478]
Seems a free cool one online Maxi *4 and / Or remove JPG artifacts !

https://deep-image.teonite.com

so (1297 * 1080 ) * 4 ;)

From: mkdm
23 Nov 2018   [#479] In reply to [#478]
Cool!

Good to know. Thanks.
From: Frenchy Pilou (PILOU)
23 Nov 2018   [#480]
Another cool free one! On Line!

https://deep-image.teonite.com/

No Register ! *2 * 3 * 4 , And / Or kill JPG artifact! ;)

And seems no limit of number images!
From: Frenchy Pilou (PILOU)
23 Nov 2018   [#481]
Speedy MagicaVoxel Workflow:

draw anything in seconds...


General Rendering...
Give one color & thickness (or not) to Edges, Grid, Ground, Background...
A material to some colors Glass to the faces's color for transparency, Metal to Ground if you want reflection


Explore the interior with a little Stereocopic vew ... that are infinite possibilities!
you can apply some color effects in your favorite 2D prog MV or others
In less 3 minutes you will have amazing architectures views! :)

From: Frenchy Pilou (PILOU)
24 Nov 2018   [#482]
For keep the speed training day! :)


From: Frenchy Pilou (PILOU)
24 Nov 2018   [#483]
About Speed modeling :)
Create a simple and complex Pattern at the same time, save it. And with each new object, use it in aTtach mode (with Shift when you want to "cut" to the mass). Coupled with the "internal" displacement of the matrix (CTRL), it is extremely effective! (and 3 keys to remember, T, Shift, CTRL) for "short memories" like me, it's perfect!
And of course you can create a Pattern in a 2nd matrix and bring it back to the 1st one as you wish!



6 secondes modeling and if you have yet render parameters for previous objects that all!



Some colors effects if wanted...


From: mkdm
24 Nov 2018   [#484] In reply to [#483]
Thanks for the mini tut!

Show messages:  1-4  …  405-424  425-444  445-464  465-484  485-504  505-524  525-544  …  3005-3010