Humans are quite good at detecting motion. How about machines? Can computers automatically segment objects in videos? I dedicated my Master thesis to a particular variant of this question. Here are a few videos as a preview.

For some reasons, the front shifter of my bicycle stopped working properly. Time to take a screwdriver and see how the shifter actually works.

German keyboards are usually pretty annoying for programming. Mapping the third-level of the buttons D, F, C, V to {, }, [, ] places the curly braces and brackets at more convenient locations without destroying much of the benefits of a German keyboard.

This case study shows the effects of acceleration, top speed, deceleration on total travel time on short distances at the example of Transrapid in Shanghai. Bottom line: it is the average speed that matters, not the top speed. Groundbreaking result … no! But it is an illustrative example with concrete numbers that shows why reporting top speeds creates illusions.

Ten lines of PostScript are sufficient to draw a pretty self-similar pattern constructed by packing circles into circles.

Start small: my first Arch Linux package submitted to the AUR. luksctl manages (portable) LUKS-encrypted devices with a logical volume manager on top of it. Profiles control how the devices are (un)locked, the logical volumes (de)activated, and the filesystems (un)mounted with a single command. Details are found at luksctl on Bitbucket.

Curious: how far can we push Google’s speech recognition web service that is used in Chrome 25+ for speech input? I measured the word accuracy of the service on the TSP speech dataset that consists of recordings of the Harvard sentences in order to get the picture.

It is now a couple of years since affordable depth cameras have entered the market. Out of a developer’s perspective, the Asus Xtion Pro Live seemed to be better suited for development than the XBOX 360 Kinect. I still think so. Here are my first impressions gained when working with the Xtion in a university project.

Google powers a mostly undocumented web service for speech recognition. The web service accepts audio data and returns a transcription. Here is a way to communicate with the web service via HTTPS POST and Python.

Google Chrome 25+ features a nice extension for speech input, which is enabled by adding the x-webkit-speech to text input fields. I put it into action. How well speech input actually works depends heavily on what kind of words you use, though.

Four longitudinal chassis beams of a Chevrolet pickup truck. The nodes of their corresponding finite element model shall be partitioned into clusters such that the nodes within one cluster show similar behaviour in a row of car crash simulations. The question: is the local clustering approach (LLCA) applicable? A summary of a research project.

For those like me who cannot switch easily between different keyboard layouts: How to use the German PC keyboard layout in Mac OS? This recipe worked for me: www.nullpointer.de/stuff/keylayout

Successful life-hacker experiment: Removed all the keys from the keyboard, washed both keys and chassis, and almost as good as new. A Swiss knife is useful, but not absolutely necessary.

Time to play search engine and experiment with PageRank. How much time does it take to page-rank a network with a million pages? I got curious during working on a university exercise and tested two power iteration variants and one Monte-Carlo method.

A minimal implementation in C++ of a role-playing game with 2D graphics, as part of an exercise at university.

See a live animation of how an artificial neural network is trained to represent a Gaussian-shaped function.

Graphs and algorithms are tightly connected. One surprisingly simple way to inspect algorithms graphically is to let the algorithms plot graphs during their execution using GML (Graph Modeling Language). The produced graphs can be formatted with graph editing software.

A blog is a nice opportunity to have a crack at different technologies. Here is a summary of systems and software involved in running this blog.

A road story on how to use my HP Officejet 6500 E710n-z printer on Arch Linux without dispensing the duplex unit and the feed scanner.

Entering a TODO item into a task manager should be as easy as typing a few words and hitting enter. Moving your TODO item to another list should be exactly one mouse-click away. All lists should be shown at once. The experimental task manager “dropandforget” does just that.

Surprisingly much for the price of 299 SEK: while the range is rather limited, the router features a separate guest network, an intuitive and comparably well-designed user interface for router management, okayish help information, humble packaging, factory defaults easy to find, a short Ethernet cable, and two years guarantee. Good for private use. Quality needs to be appreciated!

Suppose, two people are working together on a set of files. How to find a version control solution that is accessible for everyone (such as Dropbox) and powerful enough for sharing code (such as git)?

Finally! Now it is possible in Munich to look up directions and connections with public transport on Google Maps (bus connections are still missing!). According to Süddeutsche Zeitung, this took MVG (the local company responsible for operating public transport) a whole man year to finish the interface to Google Maps.

German keyboards come with ÄÖÜ. But they lack the character Å, which is frequently used in Scandinavian languages. Here is a quick-and-dirty fix to customize the xkb map for X11 on Arch Linux.

Pattern recognition (EN2202) at KTH Stockholm turned out to be a very good course, concerning lecture, exercises, and the accompanying project work. Together with a fellow student, we created a rudimentary recognizer for spoken words. It is far from production quality, of course. While a good part of the coding has been prepared for the students, the task left enough freedom to play around with hidden markov models in the context of word recognition.