On 16-4-2013 19:52, Anthony D. Baye wrote:

> I like the textures, except that the stones in the street look too clean and
> evenly worn for it to be an "Old" city.

Of course. Be aware that this is all in some kind of beta state. The end 
product will show a very different street surface texture.

>
> The things that jar my sensibilities are the 2-1/2 foot drop from the bottom
> step of the second house on the right, the perfectly clean lines of the
> buildings and the grade of the street from one side to the other.

<grin> Most of this has been solved at this moment, except for the clean 
lines of the buildings. Some degrading has still to be done on them.

>
> Just a suggestion, but you might consider making the street more level, if you
> can.  I understand that it's all one height field, but maybe you can find some
> way to deform it - say you were to take a spline of the street and flatten the
> pixels within a certain distance on either side, not necessarily perfectly
> level, but more level than it is.

This has been solved. For the close-up views a separate street surface 
is generated, following the height_field but level cross-streetwise. 
This is not an automatic process but modelled separately. Not so 
difficult to do as I use a high-resolution mesh copy of the local 
height_field for modelling purposes.

> I don't know how your macro would handle the adjustment to the terrain.

Simply by making a union of the street surface object and the 
height_field before trace()

A new street view is being rendered.

>
> I really love the concept.

Thank you indeed. I have not finished exploring the possibilities.

Thomas

Post a reply to this message

The same street after some improvements.

- a level street surface object is added.
- code influencing the buildings along the street boundaries has been 
fine-tuned.
- doors, shutters, and bars have been added (not all visible here though).

Thomas

Post a reply to this message

Attachments:
Download 'gancaloon_old city street_07.jpg' (133 KB)

Preview of image 'gancaloon_old city street_07.jpg'

"s.day" <s.d### [at] uelacuk> wrote:
>
> Wow, thanks Alain, you have just supercharged the radiosity gathering stage of
> my renders. I need to do a comparisson for quality, I usually run a no focal
> blur/AA render to gather the radiosity data then reload this data with pretrace
> set to 1. The radiosity render was running at about 5 hours and with these
> settings is now taking about 7 minutes!
>
> Will run the final high quality trace now to see if it looks as good as my 5
> hour trace. Oh the hours I have wasted ;-)
>
> Sean

Ahh.. thought that was too good to be true, the quality was not so good, still
with a few tweaks I now have almost identical quality with a 2 hour radiosity
gathering stage so still a big improvement.

Thanks

Sean

Post a reply to this message

Thomas de Groot <tho### [at] degrootorg> wrote:
> The same street after some improvements.
>
> - a level street surface object is added.
> - code influencing the buildings along the street boundaries has been
> fine-tuned.
> - doors, shutters, and bars have been added (not all visible here though).
>
> Thomas

Hi Thomas,

Looking better, I think adding some open shutters and mabe the odd door at a
random angle would be good.

Sean

Post a reply to this message

On 17-4-2013 13:10, s.day wrote:
> Looking better, I think adding some open shutters and mabe the odd door at a
> random angle would be good.

Yes, I could indeed open some shutters. At this moment one of the two 
available doors is ajar, but unfortunately it is not one of the doors 
shown here :-) I could probably automate opening and shutting the doors 
randomly. On my ToDo list.

Thomas

Post a reply to this message

An update of the Old City overview. State-of-the-art.

Thomas

Post a reply to this message

Attachments:
Download 'gancaloon_old city overview_03.jpg' (222 KB)

Preview of image 'gancaloon_old city overview_03.jpg'

From the number of houses (single and double floored) I guestimate that 
the population lies between 25000 and 30000 in this part of the city. :-)

Thomas

Post a reply to this message

Am 15.04.2013 23:46, schrieb Alain:

> The double nearest_count can affect the quality and the speed. The first
> value set the maximum bound to be used in areas where the samples have
> more variation. The second value is the minimal value to be used in
> areas where everything is prety uniform, like a large, plain wall.

No, not exactly.

Normally, during pretrace the whole image is scanned over and over again 
with ever increasing resolution, as specified by the pretrace_start and 
pretrace_end parameters. In each pass, for each individual location 
probed it is checked whether the nearest_count condition is satisfied, 
and if not, another sample is added there.

This is rather inefficient, as in regions where the individual locations 
probed already did satisfy the nearest_count location, it is unlikely to 
find individual locations that don't satisfy that condition yet.

If you specify a second parameter to nearest_count, regions where the 
/average/ sample density satisfies this second condition in one pass are 
skipped in any further pretrace passes.


Thus, the second parameter to nearest_count can be considered an average 
sample density to aim for, while the first parameter can be considered a 
local maximum sample density.

Post a reply to this message

>Thomas de Groot  on date 17/04/2013 15.53 wrote:
> An update of the Old City overview. State-of-the-art.
>
> Thomas
>
Both the last views of Gancaloon are magnificent!
Paolo

Post a reply to this message

Thomas de Groot <tho### [at] degrootorg> wrote:
> One step further. Some code clean-up and improvements, and, as shown, an
> added random texturing of the buildings.
>
> This version of the macros is gradually diverging from the one in
> p.b.s-f. I shall see how I can include the new features in the 'public'
> version ;-)
>
> Thomas

I can't be the first one to ask if you've seen "Marwencol." Just in case, I'd
strongly suggest watching it if you haven't.

 -Shay

Post a reply to this message