Posts from SIGGRAPH

[Daniel Livingstone]

SIGGRAPH proper starts tomorrow. Meanwhile, its day 2 of the co-located conferences Computational Aesthetics (CAe), Sketch Based Interfaces and Modelling (SBIM) and Non-Photorealistic Animation and Rendering (NPAR). Site here: http://www.cl.cam.ac.uk/conference/cae-sbim-npar-2011/

 

The NPAR work is probably the most interesting (for me anyway), including a range of work that create NPAR images and video output (including real time rendering) from 2D and 3D. 3D non-photorealistic rendering is especially nice for creating 3D games with highly stylised graphics - as the photo-realistic arms costs increase year on year (this year LA Noire and Crytek 3 engine set the high bar, next year they mark the standard...), stylized visuals can reduce costs while allowing games to set more distinctive art styles.

 

This conference also has a nice mix of artists (industry and non-commercial), researchers and industry.. and the talks accordingly come from a range of perspectives - some very technical, some very conceptual.

 

Favourite talks so far:

Immersion and Embedding of Self-Crossing Loops - turning a 2d sketch of a shape (which crosses over itself) into a 3d form

Generating Op Art Lines - turning 2 and 3 colour images into op-art style images (see here: http://en.wikipedia.org/wiki/Op_art )

Snaxels on a Plane - snaxel methods are one set of techniques for creating 2d contours from 3d shapes. This paper detailed a method and algorithms to achieve this, and a range of rendering techniques. Able to create highly stylised 2d images from 3d scenes

A Sketch-Based System for Highway Design - add roads to a 3d terrain simply by sketching a line: generated roads can be generated from the sketched line according to rules governing allowable gradients, tightness of curves etc.

Aesthetic Agents: Swarm-based Non-photorealistic Rendering using Multiple Images - using AI to create art - some very nice images and output. As with the OpArt paper, this was artistically driven - not very applicable in game dev, but some very nice outputs.

 

Lots of these papers could potentially be used as start points for honours projects (or indeed PhDs...), and some of this work might be turning up in AGP this year. You can see the paper abstracts online at the conference website, and full papers should be available on the ACM digital library soon (available on campus)

 

I also had a chat with some folk from Blizzard and can reveal that for anyone wanting a programming dev job with Blizzard, there are two basic routes:

1. Get a programming job somewhere else in industry and apply after you have a couple of years experience

2. Get a job in another area of Blizzard (games mastering/QA) while keeping up programming on the side. Then after a year or two apply for a programming job internally

The third route is to get an internship in California - internships are paid, though you will obviously need to raise enough money to get yourself out to the states and to cover your first few weeks.

 

 

 

 

[Daniel Livingstone]

Never did manage to post about the final day of SIGGRAPH... so here goes...

In the morning I attended a Studio session on lighting in Unity. Basically learned, that Unity works best if you limit the number of point lights that can cast dynamic shadows. Unity will build shadow maps based on the active lights, and the performance of this takes a big hit if you have multiple point lights in your scene.

There was some discussion on deferred lighting in Unity, which can be used to improve performance in many situations where you have e.g. larger numbers of spot or partially occluded lights - but does not work too well with transparent objects, requiring an extra render pass to deal with these. If you are finding rendering too slow in Unity, you might want to check for one feature: objects have a tick box to indicate that they are capable of casting shadows. Turn this off for small objects, and see if things get better.

 

After some more time at the Expo, I dropped back in for part of the course on Filtering Approaches for Real Time Anti-Aliasing. I actually only stayed for the beginning sections of this, which was good to get me up to speed with the basics. I left before they got to the latest brain twisting techniques and methods.

 

Then off to one of the closing talks... on fluid dynamics and particle effects in PixelJunk Shooter 2. This was generally very specific to 2D particle effects - their games are coded to work with a known resolution on the PS3, using a grid of particles that matches the resolution of the game. As such, they are able to adapt their particle system for a wide range of uses. Nice talk, but not terribly easy to adapt for 3D.

 

Now SIGGRAPH is over, I have two DVDs full of conference materials to try and absorb over the coming months...

 

I did spot that SIGGRAPH uses an army of volunteers - these volunteers get free passes to the full conference, and can even apply for a travel bursary. If you are serious about graphics, get yourself to SIGGRAPH - next year the event will be in LA. Start saving now ;-)

[Daniel Livingstone]

Trying to remember what I did today... it is all starting to blur!

 

- Attended the OpenGL session where 4.2 was formally announced (only 2 days after it was announced online!). Were some good presentations on different aspects of OpenGL. Splash Damage discussed the techniques used in Brink (the PC version uses OpenGL). They are currently recruiting just now... just saying. I also won a AMD usb stick for being the first to answer the question "What is the shortest function name in OpenGL compatibility mode?". If I'd been first to get one of the other questions I might have won a super duper graphics card... ho well.

- InformIT are also selling books cheap. Order from informit.com/siggraph and use the code "SIGGRAPH2011" at checkout to save 35% on purchase price

- AMD, NVIDIA, Khronos (opengl.org) and Intel all either have their SIGGRAPH material online now or will do in the next day or two. Check the respective developer sites.

- gDebugger, the debugging tool for OpenGL, was bought some time ago by AMD (not sure when), and is now available for download from http://developer.amd.com/tools/gDEBugger/Pages/default.aspx It seems to be free now, but it might not work for NVIDIA graphics cards now (requirements include AMD Catalyst graphics drivers!). gDebugger also now works as an OpenCL debugger.

- The OpenGL 4.2 Reference Card is now 12 pages long!

[Daniel Livingstone]

Interested in research (or honours projects) in areas of narrative or modelling or understanding players in games? Lots of good work out there, current talk led me to the papers available here from the Intelligent Narrative Computing group at Georgia Tech:

https://research.cc.gatech.edu/inc/publications

[Daniel Livingstone]

More posts to follow, in the meantime some useful links from various talks, handouts and events...

 

• NVIDIA Parallel Nsight brings GPU Computing into Microsoft Visual Studio. Debug, profile and analyze GPGPU or graphics applications using CUDA C, OpenCL, DirectCompute, Direct3D, and OpenGL. http://developer.nvidia.com/nvidia-parallel-nsight
• Courses in graphics from Stanford. A range of courses, from basic to very advanced. http://graphics.stanford.edu/courses/
• Mali developer - learn about ARMs own GPU platform. www.malideveloper.com
• SIGGRAPH education resources http://education.siggraph.org/resources
• Cheap books! Save 25% on Wiley books by using the SIGGRAPH promotion code "SGH11" when you order books from www.wiley.com/go/press (until Sept 1)
  Save 15% on CRC Press books until the 9th of September using the promotion code "XHK10"
• Free access to graphics research papers - for the next 10 days a limited selection of papers from SpringerLink from journals such as 3D research and Virtual Reality will be free to access (marked by green box on contents list) www.springerlink.com
• ... and I was also able to chat with a couple of companies about opportunities for student projects, internships and vacancies for full time posts. Hopefully more information to follow!

[Daniel Livingstone]

ps last year's Beyond Programmable Shading materials are online here: http://bps10.idav.ucdavis.edu/

 

[Daniel Livingstone]

Yesterday I spent most of the day in the Advanced Rendering Techniques for Games session. The morning half of this was tough going in places, with the speakers assuming familiarity with a number of methods, techniques and terms that I hadn't really encountered before. Luckily Part II was easier going that Part I, but in both I had to duck out at points to save myself from neural meltdown.

The slides and materials are not available yet - but keep an eye on http://advances.realtimerendering.com/ for their appearance. While you wait, get familiar with basic HDR rendering techniques and terms - and the use of environment probes to support environment mapping in large levels (basically multiple environment maps!)

 

Now today I'm in the Beyond Programmable Shading course... this is the 4th year the course has run, and the opening speaker is noting that it is a very fast changing field, with radically changed notes each year. Hopefully there are some good fundamental principles that can be extracted - otherwise there wont be much point teaching stuff that changes every year (it would be out of date by the time anyone graduated!)

[Daniel Livingstone]

Copies of papers from SIGGRAPH for those without access to the ACM digital library:

http://www.realtimerendering.com/blog/some-info-on-the-siggraph-201...

[Daniel Livingstone]

ps if anyone is interested in the full siggraph schedule (just to see what is being talked about), you can see the timetable in handy google calendar form here:

http://skitten.org/blog/2011/06/12/siggraph-2011-google-calendars/

[Daniel Livingstone]

The OpenGL tutorial (see link below, also Ed Angel's page at http://www.cs.unm.edu/~angel/ for additional materials from the latest edition of the Interactive Computer Graphics book) was a good summary of the current state of OpenGL - including a few brief words on Geometry Shaders and Tesselation. Highlights are probably the examples showing shader implementations to replace the fixed function pipeline's Phong Shading model for lighting - plus a couple of additional examples showing how some very minor additional tweaks can dramatically stylize the output.

Good material but very dense - if you have already done the AGP module, the tutorial should be a worthwhile condensed refresher. If you have done RT3D but not AGP yet, then the tutorial might help you prepare for what is coming! Do check it out!!!

[Daniel Livingstone]

A nice one for any android owners wanting to play with shaders:

https://market.android.com/details?id=com.simongreen.shadertoy