A 26 Giga pixel panorama

Technical characteristics

The picture was made with the Canon 5D mark II and a 400mm-lens. It consists of 1.665 full format pictures with 21.4 megapixel, which was recorded by a photo-robot in 172 minutes. The converting of 102 GB raw data by a computer with a main memory cache of 48 GB and 16 processors took 94 hours. With a resolution of 297.500 x 87.500 pixel (26 gigapixel) the picture is the largest in the world. (stand December 2009)

Text and panorama here : link Thanks to Dlpnet. 

It's not a Full sphere panorama... neither a full cylindrical... but it is nice and big...

3D interaction with Surface

This video present a nice concept to have stacking object that can be interpreted in by surface (pattern analysis by the camera through the glass).

Each element have a lens that can show the circular code of the bottom piece. Simple idea but cool.

I let you see the video :

Source : link

Visual System on GPU

The MIT and Harvard researchers try to simulate a Visual System with GPU. They demonstrate a better way for computers to ‘see’ through encoding the Visual System physiology (Brain area V1,V2,V4) as black box encoded numerically.

It remember me my engineer courses on Visual System physiology... 

So the GPU "Eye" is bigger than ours... and composed of 16 GPU...

A video explaning the biological stuff conversion to machine engineering...

Finding a better way for computers to "see" from Cox Lab @ Rowland Institute on Vimeo.

Source : link

Tiny projector for fun and for art

I definitively love this art stuff.

It's a nice use of camera with built-in  tiny projector.

The idea is cool and the show is very fun. The big tower real time performance show is neat !

C++ reference, C reference

How obtain a good overview of C and C++ native library. The response is simple search information on good website :

cplusplus Purpose an overview with sample of how use the C/STL functions.

http://cplusplus.com/reference/

You have excellent graph that show the dependencies of each STL module. A forum to ask question, It's pretty cool !

cppreference

This website cannot replace the first one but come in complement. You can see the references in your language if already people have contributed.. It's pretty nice.

Also you have a guide that come in complement to cplusplus.com you have a section General Topics that allow to cover the subjects: Pre-processor commands, Operator Precedence, Escape Sequences, ASCII Chart, Data Types,Keywords, Header Files.

So if you do not know what you can choose between use deque, vector, list, or use which algorithm to perform sort/find operation in array do not hesitate, take a look to this website. You will become a more effective programmer by using the good tools and the algorithm that fit the best the data. ( set container is very usefull after you know for what it is done !).

ABB FLexPicker robots

ABB Flexpicker robot are fast and they allow to pick and move somethings in a pretty fast way.

By look at the following video we can see that a basic computer vision task allow to find the objetcs, and after with a queue algorithm the robots make a choice to move all the cakes in the boxes...

Pretty good.

US Air force build a PS3 cluster

The US Air Force have already a PS3 (old generation) computing cluster of 336 consoles. But they want extend it with up to 2,200 new PlayStation 3 consoles for research into supercomputing.

It seems that they perform some "processing multiple radar images into higher resolution composite images (known as synthetic aperture radar image formation), high-def video processing, and 'neuromorphic computing'.

Source : link

Cell Processor spec : link

Sonnet for Lena

O dear Lena, your beauty is so vast
It is hard sometimes to describe it fast.
I thought the entire world I would impress
If only your portrait I could compress.
Alas! First when I tried to use VQ
I found that your cheeks belong to only you.
Your silky hair contains a thousand lines
Hard to match with sums of discrete cosines.
And for your lips, sensual and tactual
Thirteen Crays found not the proper fractal.
And while these setbacks are all quite severe
I might have fixed them with hacks here or there
But when filters took sparkle from your eyes

I said, "Heck with it. I'll just digitize."

.Thomas Colthurst.

Source : Here 

Sony plans for 3D support on PS3

Sony is now letting everyone know that the PlayStation 3 will factor into the company’s overall 3D television strategy. How, you might ask, is the PlayStation 3 going to factor into their 3D television strategy?

Well, with a firmware update, of course! All PlayStation 3 consoles will be firmware upgradeable to support 3D televisions in the future. Of course, no one is talking about the fact that besides needing a 3D television to hook your PlayStation 3 to, you also need software content to take advantage of the technology.

The news that Sony is planning 3D support for the PlayStation 3 is an interesting development.

Source FudZilla

My opinion : It will be hard to Sony to support all 3D television, because if they want use auto-stereoscopic display they have many one with different capacity (in term of the number of viewpoint displayed and the user must be at a given distance to see it correctly... it impose to the user to have a correct setup to have the best result...) So, Wait and see !!

Or another possibility is that they will sell 3D glasses (red/green, electronic shutter...)

Google Image Swirl, keywords based image clustering

Google have launch a new service. A web clustering exploration. The techno is based on keyword. All the image that share a similar keyword are linked. The interface is nice and powered by flash. But the keywords approach show rapidly is limit, the image do not share at all visual content (sometimes).

For exemple type Paris (you will find Paris hilton and Paris city photos....)

On the following experiment you see that I have typed 'Paris'. I have navigate through Paris Hilton clusters and finally I have found a architectural content... (From Paris city, Sainte Chapelle ).

So the approach is nice, but based on keywords seems to be very limited. But is allow to have a nice filtering of information.

CUDA Toolkit 3.0 beta released, now with public downloads

Nvidia have released to public the CUDA Toolkit 3.0.

I will not list all the Highlight.... they list because is too long.

But the most important point (related to image processing and OpenGL application are the following points) Support for all the OpenCL features in the latest R195.39 beta driver:
Double Precision
OpenGL Interoperability, for interactive high performance visualization
OpenCL Images support, for better/faster image filtering
32-bit Atomics for fast, convenient data manipulation
Byte Addressable Stores, for faster video/image processing and compression algorithms
Support for the latest OpenCL spec revision 48 and latest official Khronos OpenCL headers as of 11/1/2009

OpenCL Tutorials

AMD/ATI have a great page related to his "ATI Stream" tech. You can have access to "Developer Articles & Publications.

Notice that you have two tutorial on OPENCL.

Image convolution,

N-Body Simulation.

I'm quite impatient to test OpenCL for real.

Contour Tracing

In Image processing you have many ways to draw the contour of shapes. You can compute the contour yourself and display it (with morphologic contour (dilated shape - normal shape) ), but you can do it also with connectivity (by go along the edges of the shape).

The link here explain the different state of the art approach.

Octave/Matlab graphics Master

You will find here a powerful guide to make graphics/plot in octave matlab.

http://iste.epfl.ch/cours_matlab/graphiques.html

I know it's written in french, but a picture with universal source code is better than thousand words !

All the following list is explained :

Pie/Bar/Histogram/Stem//Plot/Polar/Meshgrid/Quiver (gradient field) / Slice (3d data).... 

Go for it ?

Google just introduced a new open-source programming language: Go. The language tag line is simple, fast, safe, concurrent, fun and open-source. Would you give it a try ?

3D points to continuous surface

Sparse information are often the result of computer science task because we have found robust information for only a part of the dataset. But we want to reconstruct a credible result that can represent the best the sampled data in front of the real data.

Interpolation is a good way to fill holes. We all know method like Bezier, Lagrange, B-Splines, Nurbs methods. But other method exist and I want mention them here :

Surface reconstruction algorithm gives us a much better feel for the data. There are Kriging that perform this task, but Kriging can be slower on large data sets. I purpose you to read the master thesis of Mike Thiesen. A FAST LAYERED ALTERNATIVE TO KRIGING.

Go ahead, make something Unreal. Free Unreal Dev Kit

What : UDK is Unreal Engine 3 – the complete professional development framework. All the tools you need to create great games, advanced visualizations and detailed 3D simulations. The best tools in the industry are in your hands.

Who's it for  : Anyone. Everyone. You. Unreal Engine 3 has been used by game developers, researchers, television studios, machinima directors, artists and students. If you have an idea that needs to be brought to life in a game engine, UDK is for you.

Why : Sample are provided, it's free for not commercial use. (Make your levels for free and test them).

Make concept game with your team on a very good engine. I just want to add : "You are The next-gen".

You must take a look to Feature, and Showcase webpages.

A breast implant simulator

You do not dream, a french company called NEUROSYSTEMS created a software called Simplant that is able to simulate breast enlargment in 3D.

You can see here some video of the simulated torso (here).

My regret, is that the torso is not modelized from photos but from measurement of the studied body.

Ubuntu 9 3D desktop

The 3D Desktop of Ubuntu Karmic Koala edition. It's nice to see new concept to handle multiple workspace (virtual desktop) on a single screen. It use compiz (transparency, sphere, cube, cylinder...).

Augmented reality in train station

I have see in the Paris (Gare de Lyon) train station a augmented reality application.

Basically it show the shops, train places, train station reception desk. The screen is touch sensitive and you have to turn the screen in the indicated direction and a video show you the path you have to follow to go to the desired place.

It works not so bad. The system is called "DirectiVision". I cannot know more about the company for the moment because the website is under construction. (a french company).

The system do not use computer vision system (I think), only a rotation sensor to know in which direction the user look at.

Computer Vision/ Graphics Papers

If you find a paper of a given computer vision conference you have to check this list :

http://www.gmazars.info/conf/index.html

You will find the papers of :

• ICCV 2009
• CVPR 2009
• ICPR 2008
• ECCV 2008
• CVPR 2008
• ACCV 2007
• ICCV 2007
• CVPR 2007
• ECCV 2006

They are sorted by categories. (Tips : subscribe to the RSS stream to be notified of the new updates).

Note for the graphics guys you can find your happiness here : http://kesen.huang.googlepages.com/

DX11 Tech preview "Heaven Benchmark"

The Unigine team have released a new tech demo called "Heaven Benchmark" to show the advantage of using DX11 for Hardware tessellation purpose.

The following features are demonstrated:

Support of DirectX 9, DirectX 10, DirectX 11 and OpenGL

Comprehensive use of tessellation technology

Advanced SSAO (screen-space ambient occlusion)

Volumetric cumulonimbus clouds generated by a physically accurate algorithm

Dynamic sky with light scattering

Interactive experience with fly/walk-through modes

Benchmark mode

 The video of the demo:

To see in real time the Hardware tessellation you can see this video (Tessellation On/Off) :

Hardware tessellation look promising and give an excellent way of have a nice vertex map to fit the best to be adapted to the height map and so give the best visual result (compared to Parallax mapping).