Printable Version of Topic

Click here to view this topic in its original format

Unmanned Spaceflight.com _ MRO 2005 _ HiRise in 3D

Posted by: Sajid May 12 2010, 09:44 PM

Hi guys,

I thought you might find this interesting:
http://www.youtube.com/watch?v=35CHx3BHvd0
http://www.youtube.com/watch?v=s2Pw2MdtWXQ

Its a real-time rendering from my algorithm, of a DTM of the Melas Chasma basin. Its a 3000x9000 crop of the original data rendered at native resolution (equivalent to 27 million screen-aligned polygons), without polygonization, compression, or data loss.

Kind regards,
Sajid

Posted by: Sajid May 13 2010, 08:01 PM

132 views and not a single comment?

Hmm...I thought there would be more interest in a full-res real-time viewer around here...

Posted by: Stu May 13 2010, 08:10 PM

Don't worry; one thing you'll learn here is that people are very busy, and often don't have time to comment on everything they see, even if they really enjoy or appreciate it. No comment does not equal no interest or no appreciation. If I commented on everything I was linked to here I'd have no time for sleep. Likewise I don't expect comments in response to things I post.

Posted by: Sajid May 13 2010, 08:46 PM

Ah, I see. I'll have to be more patient.

There is also a possibility that people don't quite get the significance of the fact that its not a pre-rendered video, but a demo of interactively exploring very high-resolution planetary data, without having to convert it to a polygon-mesh (which usually loses quite a bit of data). Its as close as we will possibly get to exploring the real thing via the DEM images we have at the moment.

Posted by: Nirgal May 14 2010, 09:04 PM

That's very impressive how you and Adrian (Mars3D) are doing such things in realtime !

The rather simple renderer I wrote (without GPU support) for the visualization of my photoclinoemtric high-res DEMs takes dozends of seconds to render a single frame of a HiRISE DEM at full 25cm resolution with a billion triangles ... no way I could be doing this in real-time.

Right now I too am thinking about a faster renderer (based on clipmaps, ray-casting, GPU-support etc.) but unfortunately don't have enough time ... I'm still too busy working on the construction algorithms for the DEMs itself ...

One question: does your renderer also support global lighting with (soft) shadows ?

Anyway: really great work, keep it up !

Posted by: PDP8E May 15 2010, 01:16 AM

Hey Doc,
the impressive part was the real time response.
...do you have to water cool that cpu??
<nice work!>

Cheers

Posted by: Sajid May 15 2010, 02:38 AM

Thanks for the kind replies.

@PDP8E: Actually its the algorithm rather than the GPU, so it doesn't get any hotter than it would doing polygons etc (probably runs cooler than doing polygons because of the lower memory load!). The GPU is just a plain old GTX 8800 (with SLI off).

@Nirgal: I haven't implemented higher-level features like dynamic lighting yet (though its own my to-do list!). The renderer works at a lower level than that. Its primarily meant to provide an alternative to polygons, and a way to off-load huge amounts of data onto the GPU with much less memory than polygons would.

The algorithm mainly runs on shaders, so if you provide the normals, then you can definitely do some nifty lighting (with shadows) in real-time. I haven't tried dynamic lighting, but running dynamic fog, and level-of-detail selection etc works in real-time without any frame-drop.

In any event, if you want, I can try the renderer on your photoclinoemtric high-res DEMs. If i get it to work (depends on the data format), you can play/work with it yourself.