In 2007 (that is 3 years ago!) Blaise Aguera y Arcas, along with Microsoft, astonished us by presenting the Seadragon and Photosynth technology. It was, and still is very very impressing and embedded below.

On this years TED conference he was at it again – and – astonishing everyone again with an augmented reality “intelligent” map application. It is amazing what his team manage to develop!

The obvious competitor is Google Maps which probably has a better grip at the market. However, Google does not compete at all with the technology presented at TED – although they do have more users – and maybe an easier solution fitting more users?

Anyway – watch the talks if you haven’t yet.

The fifth and last day of the workshop – scheduled to end 12:00. Thus it was only time for one session and a finalizing keynote.

Session 4-1 was titled Spatial analysis of urban systems which Dafna Fisher-Gewirtzman started off with the presentation Visual analytical tools for environmental and urban systems – the visual openness & visual exposure model in regard to internal space layout and functionality. The research was founded in the field of architecture – and thus presented a slightly new view and applicability of GIScience. In essence what they did was to calculate the view from inside of apartments/houses and to the outer geography termed visual openness (I think). Visual exposure was the opposite – namely the view from outside surroundings to the inside of the building. When visualized and interpreted this gives an indicator of the quality of the building. For residential buildings it is not desirable to have a very large degree of visual exposure e.g. you do not want to live in a glass house next to a busy side walk. However you do want some visual exposure and you definitely want to have a certain degree of visual openness. To calculate both these metrics they had used the spatial geography and the building data. The applicability of this were for this point to inspire architecture and design students – which it worked very good at. One example shown was a “re-design” contest for several residential blocks which were at the lower end at the social scale. Altering the interior, and to some degree the exterior,  to consciously affect the visual exposure and openness in a positive way – they found several creative solutions – which seemed to be very promising. However, as noted from the audience, the feasibility of re-designing buildings at this level is not good. The cost would probably be higher than building new buildings from scratch. All in all I found the use of GIScience as a creativity enhancing tool to be very nice – and proves that GIScience is not just an analytical tool for domain experts.

Next presentation was Spatial analysis of the diffusion of photovoltaic installations in private households in Baden-Württemberg, held by Susanne Linder. In essence they calculated the way photovoltaic (solar cells) installations spread throughout a section of Germany and at which time it did. Motivation for this was a new energy law opening for regular households to feed back energy to the power net and thus get compensation for this. I didn’t find the topic particularly interesting – although probably to some degree useful. They relied mostly on Hägerstrand’s theory on spatial diffusion (1967) – which I do not know in detail.

Third out was the presentation Development of a top-down approach to calculate residential space heating demand in Baden-Württemberg. Which followed the trail of the previous presentation. A fairly standard (at least seemed to be) GIS analysis and presentation. Essentially finding the heating (not cooling) demand of residential areas.

This ended the sessions for the workshop. To end of the workshop Jean-Claude Thill held the keynote titled “TBD” (to be decided) – as a “joke”. However the topic covered was his thoughts on differences and similarities between Regional Science and GIScience. Which was ok, but not actually any bombshell to end the workshop on.

And that was it for this time. All in all, I found the summer school and particularly the workshop to be inspiring – at least it is nice to see what others are doing – and how they do it. Of course it is also very nice to meet other people working in more or less the same field – which (unfortunately) do not happen that often in Norway.

At day four the summer school was finished and the ICA workshop started. I would guess the participants nearly doubled.

To kick off the workshop was a keynote by Vassilis Kostakos from the University of Madeira. The title being Urban encounters of the 3rd type. The talk was on how humans regularly have encounters with each other without knowing each other. For instance commuters experience this a lot. The same people encounter each other on a regular basis but they do not know each other, nor do they interact with each other other than the spatial interaction. In Kostakos’ group they have tried to support and change this by developing a system exploiting bluetooth technology – the outcome of this is cityware. In essence they place several bluetooth sensors around the city/world and detect other bluetooth enabled devices such as mobile phones. The system thus have only an ID (which is arbitrary), a spatial position and a timestamp. By this several interesting things can be calculated. If you for instance identify yourself on facebook by linking the bluetooth ID to your name, then the system can propose and suggest friends that are not (yet) of a social kind, but more like a spatial kind.

I will not go any further on the talk in this post. However I found the idea, the system and the research approach to be very interesting and motivational! So much I’m actually eager to try something similar as one of my research projects. (ideas very appreciated).

After the inspiring keynote it was time to enter the parallel sessions. The sessions lasted one and a half hour and consisted of three presentations. The sessions ran two in parallel, so one had to choose, however they had (or it seemed to) divided the session according to topics, especially visualization VS. computational topics. Which I think is a good idea.

Session 1-1 was on Spatio-temporal modeling and analysis. The first presentation was on A space-time GIS approach to exploring clusters in large moving objects datasets (Hongbo Yu and Shih-Lung Shaw). In essence they had concentrated on historical migration data in USA and visualized this by a space-time cube. The clustering were done by creating vertical “stations” where people where stationary for a long time. This worked fairly good – however an interesting approach would have been to cluster horizontally i.e. in the spatial and not temporal dimension.

The next presentation was from a group of students from Linköping titled Collaborative GeoAnalytics applied to regional temporal data (Mikael Jern, Tobias Åström and Markus Johnsson). They had developed a software called eXplorer which aimed at linking different visualization techniques like parallel coordinates and traditional geographic maps. The software was built using a flash framework called GAV Flash developed by the visualization group at Linköping. Of course the different visualizations were interactive as well as linked. The software was aesthetically pleasing and I was truly impressed! The GAV Flash is, unfortunately, not open source, nor commercially available (i.e. closed as an oyster…) However a C# implementation is available – and I will most likely look further on it and try to use it, although its rumoured that the source code is quite messy and documentation poor.

The last presentation was, yet again, Jean-Claude Thill; Modeling functional spaces: the case of the migration space in the United States. Which was on their research and notion of different spaces. The data focused on was historical migration data in USA and thus the creation of a migration space, as opposed to the regular geographical space. I find the notion and idea of different spaces very interesting – however, I found it peculiar that the visualization metaphor used by Mr. Thill was in fact of a geographical kind. Although I do not know if any other metaphors would have been better, but I think it is worth experimenting with. Something I will probably embark on (at least I hope:)

I chose to attend session 2-1 titled: Visual analytics approach to spatial analysis and modeling which started off with yet another student from Linköping, Patrik Lundblad, with the presentation; Weather and ship data monitoring applied to Geovisual Analytics. In essence they had used the same framework mentioned earlier, namely the GAV framework, but here they used the C# implementation, and hence a desktop application. Essentially the system monitored ships worldwide, their spatial location, travel route and other relevant attributes. In addition to this they included weather data, both historical (probably some weeks) and weather forecast from various sources. The visualization used was several linked and integrated visualizations, such as a geographical map, parallel coordinates and I believe a time graph. The usability seemed to be very good and the application seemed also to be impressingly fast, calculating everything on-the-fly, although I do not know the computational power of the computer used (could be very fast/expensive). One feature I enjoyed was the ability to “query” or select a spatial area of interest, and then get only data that was/had been in that particular spatial region. Implemented by a circle with selectable radius. The session got me even more interested in the GAV framework. I was also a bit impressed over the fact that they were students, probably at a master level – NTNU should really be more encouraging towards activities like this.

Second up was a participant from the summer school who presented their work on Examining statistical segmentation of multibeam backscatter images with Geovisual Analytics. As far as I understood the work consisted of mostly computational clustering and analysis of multi-beam echo sound data. More understandable a remote sensing of the seabed surface and the analysis of this. They used self-organizing maps, which I have commented before, to cluster the data and find similarities and dis-similarities. I didn’t find the presentation to be very interesting – probably due to my lack of interest of the topic. I do not enjoy remote sensing, nor self-organizing maps. However it may have been interesting for those interested in the topics.

To wrap up the second session Gennady Andrienko presented their work on Visual analytics for geographic analysis, exemplified by different type of movement data. Which was more or less a repetition of the session presented earlier this week on the summer school (and commented earlier on this blog). However he mentioned they were working on a clustering algorithm which is scalable and visually driven. Which I found quite impressing. Although I did not investigate this any further (yet).

Session 3-1; Visual approach to geospatial analysis and modeling started of with Daniel Griffith on Visualizing analytical spatial autocorrelation components latent in spatial interaction data: an eigenvector spatial filter approach. (puh). Essentially no visualization at all, but a thorough walk-through of the (complex) mathematical foundation of the computation. Which, well, I didn’t find that interesting, probably due to the limited understanding of this topic.

The next presentation was from the University of Japan Visualization of district ecological network at urban partitions for public involvement. Unfortunately the presenter lacked sufficient skills in English, which made it very hard to comprehend what the topic was about. I do think it was some kind of participatory mapping which made up a knowledge base of environmental data (instances) linked to a rule base. Although this I’m not sure at.

Ending the third session was research from Switzerland Swiss metropole: analysis and geovisualization of population and service clustering. Essentially they had clustered population data in Switzerland to automatically find urban areas. The clustering algorithm was fairly simple, using a pre-defined, rectangular, spatial grid and sort of a nearest neighbour algorithm to find clusters, very similar to standard operation on raster data. It should be noted that the population data was quite noisy, which in essence makes it more difficult, as noise needs to be filtered out or enhanced. What they found was that the resolution of the grid made huge differences in the end result, not very surprising. When tuning the resolution they managed to get the same result as the purely statistical (semi-manual) approach used by the government. I found the presentation to be ok – not very impressing, but that shouldn’t be a requirement for all kinds of research. However the presenter was asked some fairly “rough” questions criticizing their work – some with validity and some seemed to just criticize just to criticize. Which isn’t fair.

In the evening it was a workshop dinner, surprisingly at a Greek restaurant (as opposed to a Swedish), but the food was good – so no complaining:)