I've been using Subversion for several years now, but I was never really happy with it. The potential for problems when updating code on different machines or when several people work together in a loose way is very high. I am also about to release a program we are working on as open source, and would like to do that in a way that encourages people to mess with it, rather than put lots of roadblocks in their way (like having to sign up for sourceforge, getting permission from me to be able to contribute, etc.).

Also, I want to encourage my students to check in their work more often to have a more fine-grained history of changes, and be able to do that when they're not connected to the net or only through our WLAN, which blocks most ports, including SVN. Since we are using Eclipse, and some of my students are not very comfortable with using the command line, integration with that particular IDE is very important.

So I came across git and started reading, and it sounded great. A very efficient model, in use for the Linux kernel (which means it must be stable), and a storage system that made a lot of sense to me. The problems started when I wanted to actually use it. The program itself has an installer for the Mac and packages for Linux, but Windows is only supported through Cygwin (and they essentially warn people not to use it). Apart from the very technical documentation about how it is implemented, I did not find a clear explanation how to actually use it for practical work. An Eclipse plugin is in the works, but it's only available as source, and it didn't even compile on my machine. The git web interface is also annoying and hard to navigate (another hosting service is github, which is a huge improvement).

When looking for more information, I found Bazaar. Bazaar (or bzr) has is based on very similar model but was clearly developed with users in mind. It is used by the Ubuntu project, and follows many of the same ideas that have made Ubuntu so successful. There is ample documentation (a short tutorial, a long tutorial, a user guide, development guides, etc.) included with the installation package. There are installers not only for the Mac, but also for Windows, and there is an Eclipse plugin that is still alpha, but that works quite well (and it can be installed using the standard Eclipse way, rather than having to compile, export, and then add manually). There is also a hosting service for Bazaar projects called Launchpad, which is very comprehensive: in addition to code hosting, there is bug tracking, translations, discussions, and feature to-dos called blueprints.

Bazaar's command-line user interface is quite easy to learn and makes a lot of sense. This is in contrast to git, which has around 150 commands that may all be very powerful, but that are frankly impossible to figure out. When introducing Linus Torvalds (who wrote git) for a talk at Google, the introducer called git "a version control system only he can understand" - well put!

While I have found no mention of this in the git documentation (that really doesn't talk a lot about usage), Bazaar supports a number of different collaboration styles, from the single developer who keeps his files locally to centralized svn-style repositories to completely decentralized open-source style development - as well as many hybrids. Another thing that works for my students and me in particular is that Bazaar uses sftp for communication with a repository (it also has its own server and protocol, but use of that is currently discouraged), which works well over restrictive WLANs.

So while I won't argue about the technical merits, I think that what I have described shows how important documentation and support are for the success of a system. Both work very well and are in use with large software projects, but for somebody who wants to use them without caring about the nitty gritty, documentation, user interface, and support make all the difference. With the backing of Ubuntu and Canonical, Bazaar is poised to become the most used distributed source control system once people start realizing what they're missing with Subversion.

Some of the articles on this website are quite popular and receive a steady 100 visits per day, so doing away with them entirely seemed like a bad decision. So the more popular and relevant ones have been slightly updated but relegated to the archive, where they will stay.

Some articles, especially the ones on language, have been deleted and will be replaced by a series of blog postings to come over the next months. The old blog is also gone, as it was deemed of very low quality and interest. Instead, the blog is now integrated into the site, and updates will be either blog postings (which will be generally short and of a more personal nature) or articles (longer and more general).

The topics of this website will largely stay the same, but the scope and amount of material will be greatly increased. There will be more on my research in information visualization (though look at EagerEyes.org for more on that), photography, language, and other topics that interest me (like Lego Mindstorms NXT robots).

What is also new is that most pages will allow visitors to comment. I received a steady stream of emails even from my old, horrible website, so I hope that this feature will prove to be useful. Also, feel free to use the contact form to let me know if things don't work the way they should.

And finally, the design has been updated to look more clean and calm than what it was like before. I hope to have made the site a lot more readable, and also created a bit of a visual identity that ties this site and eagereyes.org together. The design will continue to be tweaked and updated.

Making abstract numbers visible is not a new idea. Everybody knows line, bar, and pie charts. These give you a better overview of the data, because they make it possible to see trends or relationships immediately. It takes a very long time to read a column of numbers, and even longer to decide the overall trend of these numbers. Looking at a simple line chart, you can tell immediately. Such qualitative impressions of the data (e.g., does the trend point up or down, which company has the largest market share and how do the others compare, etc.) are often much more useful than the exact numbers. Of course, charts don't replace the numbers, they just show them in a way that is easy to grasp.

InfoVis goes into a similar direction, but goes much further. It allows you to look at millions of data items at the same time, and to interact with it. It thus makes it possible to visually analyze your data, not just draw a pretty graph of the results that you obtained by statistical (or other) means.

An Example

This example uses a dataset about car models from 1970-82. It contains about 380 records, and about 10 values per record. For this example, only five values are used: MPG (miles per gallon), cylinders, horsepower, weight (in kg), year (two digits). The following table shows you the first five records:

The following image shows the dataset using a visualization method called parallel coordinates (click it to get a bigger image). Imagine the following steps for constructing the image from the table: replace each column by a vertical line, which represents the whole range of values for "its" column. Then, for each row of the table (i.e., each record), draw a point on each of the axes representing its values, and connect all the points belonging to the same record with lines. The result looks like this:

At first glance, this is just some line chaos. But when you look closer, you can see some structure already: there are axes with many different values, and axes with just a few. The cylinder axis only has five different values on it, and the year axis has a thirteen. For the others, you can get an idea of the distribution of the values (even though this is not very accurate, because lines can cover other lines). Especially on the MPG axis, you can see three large groups of values that seem to correspond with certain values on the cylinders axis.

But what makes parallel coordinates (and, in fact, most InfoVis techniques) useful, is interaction. You can zoom into parts of some of the axes, rearrange them, throw some of them out and bring other information in (e.g., the country of origin). Perhaps the most useful and most direct interaction is called brushing. The idea is that you mark certain values as interesting, and then look for other properties of the selected data. The following images show the results of brushing the above data set based on the year. In the left image, cars that were introduced from 1970-72 are brushed, while in the right image, the years brushed are 1980-82.

Brushing 1970-72

Brushing 1980-82

Even at first glance, the two images look quite different. On closer inspection, several interesting facts can be seen in the data. First, in the 70s (left image), the weight of cars was spread over a much wider range than in the 80s (right image): cars in the 80s were in the lower half of the weight range of the 70s. The same is also true of the engine power (horsepower axis). Looking at the MPG scale, you can also see that cars in the 70s had a much lower mileage than in the 80s (for Europeans: low values are bad here, because the MPG gives you the number of miles you can drive with one gallon of gas, as opposed to the amount of gas the car uses per kilometer).

An interesting detail is that in the 1980-82 range, there was only one car model with eight cylinders. If you follow the line from the cylinders to the horsepower axis, you can see that there is another line leading to the same value. Following that line back to the cylinders axis, we find a four-cylinder car. So the last remaining eight-cylinder had only as much power as one of the four-cylinders, and definitely needed more gas than that car (this is not really visible in this image without some more interaction). The eight-cylinder was also much heavier than the four-cylinder of the same power (this, again would need more interaction).

This was just a very simple example, but (hopefully) one that was easy to follow. InfoVis can in fact do much more, with larger data, more dimensions, and higher data complexity.

Why is InfoVis interesting?

InfoVis brings together several interesting aspects. First of all, it is graphical. That in itself is much more interesting than statistics ;). And that also means that much of the research and experience from perceptual psychology can be used to understand why some visualization methods are better than others. Examples are the Gestalt laws and preattentive vision: we see objects as groups and in certain constellations because of these phenomena.

Another field that is relevant to InfoVis are the visual arts. Visual communication was not invented by InfoVis people, and we certainly can learn a lot about how to use colors, etc. Some work has already been done in building new visualization methods on ideas from the arts, like the layering used in oil painting. Much more can still be done, though.

All this is not to say that InfoVis does not also pose technical challenges - even though this is a particularly weak spot of most InfoVis research (quite in contrary to volume and flow visualization). Especially when dealing with large datasets on the order of magnitude of one million items, it becomes crucial to design systems with speed in mind. To be really useful, InfoVis methods need to respond quickly to user input - only then, interactivity makes sense.

Conclusions

InfoVis is a fascinating application of computer graphics to real-world problems and data. It unites artistic and technical aspects in a way that is both useful and interesting.

My interest in pinhole cameras is in the pictures, not the "technology". Experiments with building cameras out of soda cans, matchboxes, or railway cars are fun, but as long as they don't deliver interesting pictures (and not just blurry ones), I don't find them very interesting.

I wanted to build a pinhole camera that would be simple and fun to use. And with that, I mean the following:

• Use film. It's quite common to use photographic paper in pinhole cameras, and there are good reasons to do so. But I wanted to use film so that I have larger tolerances for exposure (which you really need for pinhole cameras) and can make prints much more easily. Using film is also necessary for at least the first two of the following requirements.
• Easy film advance. I wanted to be able to take several pictures without having to go to the darkroom between them. The Loch-Lomo takes eight pictures per roll of 120 film.
• Easy film change. I wanted to change films in broad daylight, again without having to find a completely dark place. This is especially important if you want to take your pinhole camera with you when travelling.
• The color option. It's fun to use color film in a pinhole camera for a change (see my pinhole pictures from Portugal on Flickr). And that is much easier with film than with photographic paper. Yes, it would be possible in principle to use Ilfochrome or even color negative paper, but that simply wasn't an option for me. It is so much easier to get the film developed and printed in a lab.
• Shutter mechanism. I wanted to have a real shutter, rather than having to put a dark slide in front of the pinhole. That also makes keeping time much easier and is just plain more convenient: a standard cable release fits on the camera, which lets me take pictures from one second to many hours with great precision and convenience. 
• Resistance against wind. Building pinhole cameras from cardboard boxes is all good and well, but when you use them outside, the image is easily blurred due to wind shaking the camera. I actually had a pinhole camera made from a kit that was made from cardboard, and I had it tip over from the wind on several occasions. The weight of the Clack (which is still quite light in comparison with other cameras) is enough to keep it from doing that.

The Loch-Lomo

The name is a little play on words. "Loch," in German, means "hole" – the German word for "pinhole camera" is "Lochkamera." "Loch" is also Scottish for "lake" (you may have heard of Loch Ness). The Lomo, is a small 35mm point&shoot type camera made in Russia. It doesn't take very sharp images, and its users don't seem to be concerned with what is on their pictures.

So the name "Loch-Lomo" is kind of a joke, putting these two words together that sound similar, and that also have some kind of meaning. And if the name seems familiar, it may be because you've heard of Loch Lomond, which can be found in Scotland, Canada, the US and Australia.

The Loch-Lomo is a modified Agfa Clack, which is a box camera that was made in the 1950s and '60s. The Clack's "optical system" consists of a single lens that is able to produce a usable picture due to the fact that the back of the camera is slightly curved. I bought mine for the equivalent of about 15 US$ at a flea market in Austria (where they are quite common and cheap), to play around with 120 film a bit. But the pictures never satisfied me (they were blurry, even on the contact sheets), and so I developed a plan to modify it.

Converting the Clack into a Loch-Lomo

Please Note: I do not sell Agfa Clacks or Loch-Lomos. This page is meant to show how it's done, but you have to do it yourself. It's also much more fun this way. Believe me, it's a totally different experience to see the first image you took with a camera you built yourself.

Making a pinhole camera out of an Agfa Clack is fairly easy. First, you have to remove a screw at the bottom part of the "lens" and remove the metal strap. Then you remove the that part that has "Clack" written on it by unscrewing it. Now you can remove the lens (two screws hold the metal lensholder in place). The only thing you really want to keep is the part that has "Clack" written on it, I use it as a kind of lens cap to protect the shutter mechanism when the camera is in my photo bag. If you leave it on during exposure, you get an elliptical frame around the image, though.

Under the lens there is a piece of metal with three holes: One with a small plastic lens in it, one with a yellow filter, and one a simple hole. I simply broke that off, because I couldn't find any other way to remove it. It needs to be removed because it would cause vignetting.

Now you can see the shutter, which is mounted onto a round metal plate. It will have to be removed to put the pinhole into place, but don't break it! After removing two screws, you will find that the round plate is also attached to two wires that connect the shutter to the flash contacts on the top of the camera. I simply cut these wires, because even if I ever want to use a flash with the camera, I will have to build in my own terminals and thus will need new wires.

Between the shutter plate and the plastic body, there is a thin felt ring. Remove that carefully, so that you can later use it to make the space between the shutter and the pinhole light-proof. Now simply stick a piece of metal foil (or whatever you like to use) with a small hole in it, put the felt ring on it, and screw the shutter plate back on. Voilà, you are now the proud owner of a Loch-Lomo!

While the description might make this sound like a lot of work, this can be easily done in about 20 minutes (including making the pinhole, if you have some experience doing that). I went from the idea of doing this to having my first printed pinhole pictures in a single day: I thought about doing it in the morning, converted the camera around noon, went out to shoot my first pictures in the early afternoon, processed the film and made prints in my small darkroom in the evening.

In the picture at the top of the page, you can see a white area on the top of the camera. That is a table with corrected exposure times for long exposures that you can find on my pinhole camera page.

Making the Pinhole

I studied Medicine for about two years (when I got bored with computer science), and from that I still had a set of scalpels and blades for them. The blades come in a thick tin foil that is very nice to handle. I cut out a piece that would be big enough for the Loch-Lomo, and then made a pinhole by putting the tip of the scalpel blade on the metal and slowly rotating it. This gives a very nice, round hole and good control over the size of the hole. I set my slide caliper to 0.3mm, and compared the hole to the gap between the parts of the caliper by moving it about behind the gap, and holding both parts against the light. Any thicker aluminum or other metal foil should work, or even a piece of metal from a soda can.

The only problem I have are reflections that cause bright areas near the middle of the image. I painted the inside of the tin foil black, which reduced the problem significantly, but hasn't solved it completely.

Here is a table with the Loch-Lomo's specs.

Pictures, Experiences

Since I built the camera about five years ago, I have taken this camera with me on several travels and have been able to take some very interesting pictures with it. Changing film is very easy, and having color negative film developed and printed is quite affordable and convenient. The cable release makes it easy to time exposures from a single second (which is about the shortest possible on a sunny day with ISO100 film) to several hours. I am even considering building a little Lego robot that can precisely time exposures, especially for long exposures overnight (that need to end before it gets too light).

The long exposure times make it possible to get some very interesting shots, especially when water is involved. The image above was taken before and during a recent total lunar eclipse, with an exposure time of about 2.5 hours. It nicely shows how the moon's brightness and color change over time as it more and more of its surface enter the shadow cast by the earth.

This Pinhole Photography site owned by Robert Kosara hosted by RingSurf Next | Previous Random Site | List Sites

What is a Pinhole Camera?

The basic principle of the pinhole camera is a dark room (camera obscura in Latin) or light-tight box, with a small opening on one side to let light in which is projected onto the opposite wall. In an actual pinhole camera, this is where a piece of film or photographic paper is placed, but pinholes were and still are also used to project images like that of the sun. As early as the tenth century, experiments with pinholes were made, and in the Renaissance, the camera obscura was a popular form of entertainment and also a tool for artists to better understand perspective.

The obvious difference between the usual cameras we use and pinhole cameras is the lack of a lens. A lens gathers a lot more light on its large surface than a tiny hole, and so a pinhole camera requires much longer exposure times. It also provides almost infinite depth-of-field, though, which makes images possible that would not otherwise be.

But the most interesting part is really building such a camera. While it is possible to buy one, I would strongly advise against that. Building a pinhole camera is easy, and taking pictures with a camera you built yourself has a completely different quality than using a bought camera.

Building a Pinhole Camera

A pinhole camera can be built from almost anything. Popular objects are shoe boxes, film canisters, tin cans, and old box cameras. The main questions to answer are:

Do you have a darkroom? If yes, your options are really unlimited. If no, it makes sense to build a camera that takes roll film, which can then be processed by a lab. Also, loading and unloading the film should be possible in normal light conditions, which is not the case with most shoe box or tin can cameras, for example. A darkroom is certainly not required for building your own pinhole camera, but it does help being able to quickly process and print your first roll of film to see if there are light leaks, how well the exposure works, etc.

Film or paper? Shoe box cameras are usually loaded with a piece of photographic paper, which needs to be loaded and removed in a room that is completely dark (perhaps with a safelight). The paper is then developed into a paper negative, that needs to be copied onto another sheet of photographic paper to get a positive. This process is easy to do when you have access to a darkroom and want to do some quick experiments, but is unworkable otherwise. Paper negatives also don't have nearly the dynamic range of film, and produce very flat images with little contrast.
Film is a much better choice, but does require a more sophisticated camera. The exception here is a film canister that can be loaded with a piece of film cut from a roll, but which essentially requires the same infrastructure as using paper. Roll film is much more practical, because it is possible to take many pictures and change film in the field, rather than having to head back to the darkroom after every shot.

Once you have decided on the type of camera to build, it is easy to find materials to make it from. I made a camera out of an old Agfa Clack box camera, which I call the Loch-Lomo. A camera made from an iPhone box (which is black, and thus very well suited for this purpose), dubbed iHole is a slightly more stylish option. Justin Quinnell makes cameras out of "pocket film" (110 type) cassettes, which he uses to great effect to shoot pictures from the inside of his mouth.

A Different Perspective

The special properties of pinhole cameras make it possible to explore new perspectives that one would not otherwise have tried. The infinite depth of field and often wide angle of view make it interesting to put the camera on the ground, where it can see objects that are very close as well as ones far away.

Many pinhole cameras lack a viewfinder, and if they have one, it may not be very precise. Also, shoe boxes and film canisters don't have tripod mounts, so placing a camera precisely is difficult. Their low weight and negligible value makes it possible to put pinhole cameras where other cameras don't usually go: on top of bushes or rocks, on the ground between walking people, mounted on boomerangs, etc. 

The inherently long exposure times also produce interesting results, though similar effects can of course be achieved with conventional cameras. A camera like the Loch-Lomo has an effective f-stop of 1/256, so even in bright daylight, exposure times of less than one second are very rare. Because of reciprocity failure (see below), long exposure times become even longer, making exposures of hours, days, or months possible.

Water is especially interesting in this case, because its constant movement tends to create very pleasing effects with long exposures. Most images look as if the water was frozen, its surface looking like nacre.

Pinhole cameras provide a lot of room for experiments, and they really require experimentation. Simply making or buying a camera and trying to take the same pictures as with a regular camera won't lead to interesting results. But experimentation is what photography should really be all about!

Examples

Examples of pictures taken with pinhole cameras are easy enough to find, but one needs to be aware that one or even a set of pinhole pictures is not representative of all the things that can be done. There are different approaches, and the resulting images are quite different - just like with a regular camera!

A few examples are my own pinhole pictures on Flickr, the Worldwide Pinhole Day Gallery and pictures tagged with pinhole on Flickr, and of course Justin Quinnell's brilliant pictures.

Exposure Time Correction

The long exposure times used in pinhole cameras require some extra attention. Starting from about one second, the exposure no longer increases in a linear fashion, but needs to be corrected to account for the reciprocity failure or Schwarzschild effect. The following table gives the corrected numbers the Agfa Pan 100 (APX 100), but should work for any ISO 100 film, black&white or color (I have used these times with APX 100 and Fujicolor 100).

This table was prepared based on the following formula (thanks to Dieter Lefeling): tc = 0.1*(10*tm)^(1/(1 - lg 2)) (where tc is the corrected time, tm the measured time, and lg the base 10 logarithm).