Re: [OSM-talk] High Cartographic Quality Label Placement on OSM-based Map
Am 14.07.2013 17:08, schrieb Serge Wroclawski: On Sun, Jul 14, 2013 at 10:10 AM, Andreas Reimer andreas.rei...@geog.uni-heidelberg.de wrote: It sounds like you understand these arguments (at least partially), so let's not go over the same points needlessly, but at the same time I think it bears repeating that you are seeing various cultural biases at work (academic vs FLOSS), that the two communities are similar in some ways, but very different in others - Serge Serge, I just want to add: I fully know that people have brilliant ideas and solutions outside of academia. Most likely moreso! The scheme I outlined is not reality. It is the normative basis for funding and division of labour as viewed by government and professors. And those do provide our funding and degrees. That is all I wanted to say with that. In no way did I mean to claim academia has any monopoly of innovation. But it structrually pretends it does, and we have to live in that environment. Sorry again, Andreas ___ talk mailing list talk@openstreetmap.org http://lists.openstreetmap.org/listinfo/talk
Re: [OSM-talk] High Cartographic Quality Label Placement on OSM-based Map
Christoph, I do see your point regarding lacking information in papers. Truth to be told, most labelling research is from computational geometry and they do often not care too much about actually labelling the maps, weird as it sounds. And as a more math-based discipline, they have a very low priority on providing workable parametrisation. Exceptions exist of course, no disrespect whatsoever to all the brilliant geometers out there on whose shoulders we stand! Our paper in contrast includes all the information and parametrisation you need to implement that wherever you want. And the tileset shows you how the results will look like. Maybe we should have waited till the paper is accepted before sharing. Sorry for any confusion we caused. Thanks for your interest and time. Andreas PS: Regarding government money on research We all have limited contracts and are evaluated by journal paper output. That is the reality of it. Our boss is a strong believer in demo or die, but the general structure works against that. Am 14.07.2013 19:32, schrieb Christoph Hormann: On Sunday 14 July 2013, Andreas Reimer wrote: The whole point of algorithms research is to move beyond implementation and do research, well on algorithms, instead of software libraries. That Max has a very stable and functional framework is uncommon for the scientific community. And he built that framework mostly in his free time over the years to better test his hypotheses results. I never said anything about the whole framework. I am well aware there is much more to producing a map like the one you showed than just the label placement code. But the point is that for your work to contribute to scientific progress you need to make available everything others need to reproduce your research results. If you can do this just using pseudocode and verbal description that is fine. Since in your blog you cite Eduard Imhof who is well known for his ability to describe in minute detail his cartographic techniques i should probably give you the benefit of the doubt but knowing the complexity of label placement i do not expect this to be practical. Of course there is a lot of stuff published in science journals, especially wrt. algorithms that does not include the information necessary to reproduce the results. This does not make it right though and such publications usually make little contribution to the overall scientific progress. There are strong interests at stake that make wholesale software development at Universities a risky endeavour or plainly forbidden. You might disagree with that, but I hope you can at least acknowledge there are competing interests here which neither of us can change at the moment. Of course there are competing interests but since the financiers of public research in Europe these days do not usually require or even explicitly support making available the results to the public this competition is a little one-sided. You, the researchers working at those institutions, are the only ones who can actually make a difference here. And we hope we pushed the boundaries in what is feasible without any proprietary software a little bit. Just to clarify this - if the code is not available for others to use, modify and redistribute it is proprietary software, even if all code except your own is free. Greetings, ___ talk mailing list talk@openstreetmap.org http://lists.openstreetmap.org/listinfo/talk
Re: [OSM-talk] High Cartographic Quality Label Placement on OSM-based Map
Serge, I greatly appreciate your input from other disciplines. As to the cultural differences between fields, I was never implying the structure of either is better than the other. As much as I can see why providing the full implementation is the true proof of the pudding for some fields, in others you get shoot down for just doing engineering and wasting government money on stuff free market shall provide. I have no dog in the race, as we are GIScience people. Naturally at the crossroads of engineering and theory in any case. No matter what we do it is either too theoretic or too applied for some people. And again sorry for any implied insult, I have the utmost respect for all engineers and find the outlined division of labor myself a bit contrived sometimes. But one has to operate in the surrounding you operate in. Regarding the shoulders of giants, I do agree. That is why you cite people and relevant preceding works in papers. Thanks for your time, Andreas Am 14.07.2013 17:08, schrieb Serge Wroclawski: On Sun, Jul 14, 2013 at 10:10 AM, Andreas Reimer andreas.rei...@geog.uni-heidelberg.de wrote: Hi there, I am a colleage of Max's and we collaborated on his labelling papers. Chiming in to straighten up some potential miscommunications. I think there's general understanding but some substantial cross cultural differences coming out of this discussion. I am speaking here as someone who has worked at some academic institutions (namely NASA and NIH) where scientific research was required to be public (due to different laws, but in both cases, the law required that not only were the publications stored, but the implementation of those algorithms stored as well). But i would actually emphasize a more practical point: Since this is meant to be scientific research one can assume you publish it to allow others to independently verify your work, compare it to their own and possibly improve it - this is in the very definition of scientific work Just wanted to point out that fairly complex algorithms are actually not published the way you assume, although I totally see your point. This is certainly an understandable position from an academic standpoint, but this is where engineering (and the FLOSS world) and academic clash in ideas of how information is spread. I think that increasingly academics are trying to straddle the two worlds and finding it hard to do so. Some context: map labelling is a popular subject for computational geometry and GIScience people. And they do not usually produce or share fully-fledged frameworks, but algorithms in pseudo-code with complexity analysis in Big O notation and/or pragmatic runtime analysis as empirical exploration with a sample implementation. This may be the norm in your community, but it's certainly not the norm in all scientific communities. At NASA, certainly high level scientific papers were written regarding issues of electromagnetic reflectivity of certain wavelenghs and what that said about the composition of the chemical composition of the atmosphere- but at the same time all the implementations of those algorithms (the code) was required to be stored in order to allow other researchers to replicate the results. The implementation of an algorithm was considered part of the algorithm. Either you could consider the source code (and compiler version, library version, etc.) to be part and parcel of the algorithm, since it described the actual computational work being performed, or you could consider it simply necessary for replication and testing. Either way, both were required. The whole point of algorithms research is to move beyond implementation and do research, well on algorithms, instead of software libraries. Intentionally or not, that's insulting to those people who choose to be either creative or scientific outside of academia. Look at the work being done by various computer scientists at large commercial institutions such as Google and IBM who, after implementing their work, write it up as a scientific research paper for publication. That Max has a very stable and functional framework is uncommon for the scientific community. And he built that framework mostly in his free time over the years to better test his hypotheses results. Yes, and we're all very happy about this, and want to share in the work. Yes, once the paper is accepted, the algorithm is out there in the open and I am sure any of us will gladly answer questions and help for anyone wanting to use the algorithm. This is very good, and I'm sure I speak for everyone in mentioning that we're glad that your academic institution is not taking the stance that it owns the work, etc. Getting a set of tiles with new labelling for anyone to use is also giving back to the community already? No. This is not a giving back to the community any more than declaring a result without describing a process would be acceptable to you in an academic context.
Re: [OSM-talk] High Cartographic Quality Label Placement on OSM-based Map
Hi there, I am a colleage of Max's and we collaborated on his labelling papers. Chiming in to straighten up some potential miscommunications. Am 13.07.2013 20:36, schrieb Christoph Hormann: But i would actually emphasize a more practical point: Since this is meant to be scientific research one can assume you publish it to allow others to independently verify your work, compare it to their own and possibly improve it - this is in the very definition of scientific work. When dealing with fairly complex algorithms like here this is next to impossible to do without publishing the code. Just wanted to point out that fairly complex algorithms are actually not published the way you assume, although I totally see your point. Some context: map labelling is a popular subject for computational geometry and GIScience people. And they do not usually produce or share fully-fledged frameworks, but algorithms in pseudo-code with complexity analysis in Big O notation and/or pragmatic runtime analysis as empirical exploration with a sample implementation. The whole point of algorithms research is to move beyond implementation and do research, well on algorithms, instead of software libraries. That Max has a very stable and functional framework is uncommon for the scientific community. And he built that framework mostly in his free time over the years to better test his hypotheses results. Now i understand you might be reluctant to make available the code before a journal publication and you would not necessarily need to make it open source/free software license wise although this would be advisable as a matter of fairness when extensively using Openstreetmap data. Yes, once the paper is accepted, the algorithm is out there in the open and I am sure any of us will gladly answer questions and help for anyone wanting to use the algorithm. Getting a set of tiles with new labelling for anyone to use is also giving back to the community already? And i won't even get started on the fact that work of a publicly funded research institution should benefit the public... This is a serious point to clarify. The idea of public research is as such: - researchers push the boundaries and develop new algorithms and publish them - companies and engineers in public service use the algorithms for their products*/tasks - everything gets more efficient, more money is left for other wonderful stuff - repeat *(note who is the one making the most money with the algorithm in such a scheme) There are strong interests at stake that make wholesale software development at Universities a risky endeavour or plainly forbidden. You might disagree with that, but I hope you can at least acknowledge there are competing interests here which neither of us can change at the moment. Thanks so much for your interest in that work, I agree with you that most tile-based maps should benefit from labelling improvement. And we hope we pushed the boundaries in what is feasible without any proprietary software a little bit. Let's hope the reviewers see it likewise and we can share the work as soon as possible with everybody. Don't think we don't want to share, think of us as wanting to share very early as much as possible. All the best, Andreas Reimer Greetings, ___ talk mailing list talk@openstreetmap.org http://lists.openstreetmap.org/listinfo/talk
Re: [OSM-talk] High Cartographic Quality Label Placement on OSM-based Map
On Sun, Jul 14, 2013 at 10:10 AM, Andreas Reimer andreas.rei...@geog.uni-heidelberg.de wrote: Hi there, I am a colleage of Max's and we collaborated on his labelling papers. Chiming in to straighten up some potential miscommunications. I think there's general understanding but some substantial cross cultural differences coming out of this discussion. I am speaking here as someone who has worked at some academic institutions (namely NASA and NIH) where scientific research was required to be public (due to different laws, but in both cases, the law required that not only were the publications stored, but the implementation of those algorithms stored as well). But i would actually emphasize a more practical point: Since this is meant to be scientific research one can assume you publish it to allow others to independently verify your work, compare it to their own and possibly improve it - this is in the very definition of scientific work Just wanted to point out that fairly complex algorithms are actually not published the way you assume, although I totally see your point. This is certainly an understandable position from an academic standpoint, but this is where engineering (and the FLOSS world) and academic clash in ideas of how information is spread. I think that increasingly academics are trying to straddle the two worlds and finding it hard to do so. Some context: map labelling is a popular subject for computational geometry and GIScience people. And they do not usually produce or share fully-fledged frameworks, but algorithms in pseudo-code with complexity analysis in Big O notation and/or pragmatic runtime analysis as empirical exploration with a sample implementation. This may be the norm in your community, but it's certainly not the norm in all scientific communities. At NASA, certainly high level scientific papers were written regarding issues of electromagnetic reflectivity of certain wavelenghs and what that said about the composition of the chemical composition of the atmosphere- but at the same time all the implementations of those algorithms (the code) was required to be stored in order to allow other researchers to replicate the results. The implementation of an algorithm was considered part of the algorithm. Either you could consider the source code (and compiler version, library version, etc.) to be part and parcel of the algorithm, since it described the actual computational work being performed, or you could consider it simply necessary for replication and testing. Either way, both were required. The whole point of algorithms research is to move beyond implementation and do research, well on algorithms, instead of software libraries. Intentionally or not, that's insulting to those people who choose to be either creative or scientific outside of academia. Look at the work being done by various computer scientists at large commercial institutions such as Google and IBM who, after implementing their work, write it up as a scientific research paper for publication. That Max has a very stable and functional framework is uncommon for the scientific community. And he built that framework mostly in his free time over the years to better test his hypotheses results. Yes, and we're all very happy about this, and want to share in the work. Yes, once the paper is accepted, the algorithm is out there in the open and I am sure any of us will gladly answer questions and help for anyone wanting to use the algorithm. This is very good, and I'm sure I speak for everyone in mentioning that we're glad that your academic institution is not taking the stance that it owns the work, etc. Getting a set of tiles with new labelling for anyone to use is also giving back to the community already? No. This is not a giving back to the community any more than declaring a result without describing a process would be acceptable to you in an academic context. Max has done something great, and clearly put a great deal of time, effort, etc. into it. We're all very happy for him. But we also see that this work was not done in a vacuum, and the expectation of this community is that when standing on the shoulders of giants, you also provide a platform for others to stand on. It sounds like you understand these arguments (at least partially), so let's not go over the same points needlessly, but at the same time I think it bears repeating that you are seeing various cultural biases at work (academic vs FLOSS), that the two communities are similar in some ways, but very different in others - Serge ___ talk mailing list talk@openstreetmap.org http://lists.openstreetmap.org/listinfo/talk
Re: [OSM-talk] High Cartographic Quality Label Placement on OSM-based Map
On Sunday 14 July 2013, Andreas Reimer wrote: The whole point of algorithms research is to move beyond implementation and do research, well on algorithms, instead of software libraries. That Max has a very stable and functional framework is uncommon for the scientific community. And he built that framework mostly in his free time over the years to better test his hypotheses results. I never said anything about the whole framework. I am well aware there is much more to producing a map like the one you showed than just the label placement code. But the point is that for your work to contribute to scientific progress you need to make available everything others need to reproduce your research results. If you can do this just using pseudocode and verbal description that is fine. Since in your blog you cite Eduard Imhof who is well known for his ability to describe in minute detail his cartographic techniques i should probably give you the benefit of the doubt but knowing the complexity of label placement i do not expect this to be practical. Of course there is a lot of stuff published in science journals, especially wrt. algorithms that does not include the information necessary to reproduce the results. This does not make it right though and such publications usually make little contribution to the overall scientific progress. There are strong interests at stake that make wholesale software development at Universities a risky endeavour or plainly forbidden. You might disagree with that, but I hope you can at least acknowledge there are competing interests here which neither of us can change at the moment. Of course there are competing interests but since the financiers of public research in Europe these days do not usually require or even explicitly support making available the results to the public this competition is a little one-sided. You, the researchers working at those institutions, are the only ones who can actually make a difference here. And we hope we pushed the boundaries in what is feasible without any proprietary software a little bit. Just to clarify this - if the code is not available for others to use, modify and redistribute it is proprietary software, even if all code except your own is free. Greetings, -- Christoph Hormann http://www.imagico.de/ ___ talk mailing list talk@openstreetmap.org http://lists.openstreetmap.org/listinfo/talk
Re: [OSM-talk] High Cartographic Quality Label Placement on OSM-based Map
Hi Am 12.07.2013 20:30, schrieb Maxim Rylov: Is the algorithm available as open source? Unfortunately, the algorithm currently is not open-source, but the model that we elaborated and used will be published as a journal paper within the next few months. That's really a pitty. How can someone profit from your work (aka Shoulders of Giants)? What worth is it pointing to us, how bad our label placement is and how good it could be by not giving us the tools to fix it? Sorry for that general rant. I'm just sick of people doing wonderful things without telling anyone about how they did it. How can humanity evolve with such a mindset? Peter ___ talk mailing list talk@openstreetmap.org http://lists.openstreetmap.org/listinfo/talk
Re: [OSM-talk] High Cartographic Quality Label Placement on OSM-based Map
That's really a pitty. How can someone profit from your work (aka Shoulders of Giants)? What worth is it pointing to us, how bad our label placement is and how good it could be by not giving us the tools to fix it? I did not want to abuse anybody. At frist I tried to present the results that can be achieved with great OSM data and than tried to clarify the difference between two approaches. That algorithm is a part of my PhD research. At the moment, due to some restrictions I am not able to share the tool and algorithms with others. Sorry for that general rant. I'm just sick of people doing wonderful things without telling anyone about how they did it. How can humanity evolve with such a mindset? As I wrote before the model that we use will be described in detailes in a journal paper. One, a researcher or a developer, can benefit from it as soon as it is published. I hope that the work, we have done, becomes a part of human knowledge. ___ talk mailing list talk@openstreetmap.org http://lists.openstreetmap.org/listinfo/talk
Re: [OSM-talk] High Cartographic Quality Label Placement on OSM-based Map
On Friday 12 July 2013, Maxim Rylov wrote: As far as I know Mapnik and other open-source software take into account only the rules R1, R2 and R5 (partially, returns labels that are evenly spread out, example - http://maps.skobbler.com/on z11). And the greedy algorithm that is utilized to solve the label placement problem returns rather poor approximation to the optimum as there is no backtracking. I see - Mapnik labeling is well known to be quite bad so it is not really a good competition. But you have not said which of these rules are followed by your method. There seems to be an (only partially successful) attempt in R3 and R4 but R8 does not seem to play a role. R1 in several cases seems to work better in Mapnik. Is the algorithm available as open source? Unfortunately, the algorithm currently is not open-source, but the model that we elaborated and used will be published as a journal paper within the next few months. Well - Peter has already commented on this and i mostly agree with him. But i would actually emphasize a more practical point: Since this is meant to be scientific research one can assume you publish it to allow others to independently verify your work, compare it to their own and possibly improve it - this is in the very definition of scientific work. When dealing with fairly complex algorithms like here this is next to impossible to do without publishing the code. Now i understand you might be reluctant to make available the code before a journal publication and you would not necessarily need to make it open source/free software license wise although this would be advisable as a matter of fairness when extensively using Openstreetmap data. And i won't even get started on the fact that work of a publicly funded research institution should benefit the public... Greetings, -- Christoph Hormann http://www.imagico.de/ ___ talk mailing list talk@openstreetmap.org http://lists.openstreetmap.org/listinfo/talk
Re: [OSM-talk] High Cartographic Quality Label Placement on OSM-based Map
7/12/2013 12:18 AM, Christoph Hormann wrote: Could you explain in what ways this is the case. Since different types of labels are shown in various maps direct comparison is difficult. You seem to very well avoid overlaps between labels and none the less you are able to put quite a lot of them on the map but non-label feature do not appear to play a role in label placement and there are some strange priorities. Is the algorithm available as open source? Greetings, Avoiding of overlaps in map labeling is a constraint that is necessary. We can compare the map labeling on osm.org and on our map with the help of the list of cartographic rules for point-feature labeling. For example, they are: * R1. *Type arrangement should reflect the classification, importance and hierarchy of objects. *R2.*The lettering to the right and slightly above the symbol is prioritized. *R3.*Names of coastal settlements should be written in water. *R4.*Label should be placed completely on the land or completely on the water surface. *R5. *Names should not be too close to each other *R6.*Labels should not be excessively clustered nor evenly spread out. *R7. *Each label should be easily identified with its point-feature. Ambiguous relationships between symbols and their names must be avoided. *R8. *Label should not overlap other significant features of the cartographic background or do it as little as possible. As far as I know Mapnik and other open-source software take into account only the rules R1, R2 and R5 (partially, returns labels that are evenly spread out, example - http://maps.skobbler.com/on z11). And the greedy algorithm that is utilized to solve the label placement problem returns rather poor approximation to the optimum as there is no backtracking. Is the algorithm available as open source? Unfortunately, the algorithm currently is not open-source, but the model that we elaborated and used will be published as a journal paper within the next few months. All the best, Maxim Rylov ___ talk mailing list talk@openstreetmap.org http://lists.openstreetmap.org/listinfo/talk
[OSM-talk] High Cartographic Quality Label Placement on OSM-based Map
Hi all, We are pleased to announce that a new web map based on OSM data has just been published. You can see it on http://openmapsurfer.uni-hd.de/ (OSM Roads (New) layer). One thing that greatly distinguishes it from other online maps is the quality of map lettering. We, the researchers the researchers at GIScience Research Group http://giscience.uni-hd.de/ at Heidelberg University, have elaborated a comprehensive multi-criteria model for high cartographic quality label placement and implemented it on top of a toolkit for publishing spatial data to the web. We applied this model to render a map for the globe on lower zoom levels (z2-z12). More information about improvements and updates you can find on our Blog page http://giscienceblog.uni-hd.de/. PS: Please, take into account that the tile cache for zoom levels greater than z12 is empty. All the best, Maxim Rylov ___ talk mailing list talk@openstreetmap.org http://lists.openstreetmap.org/listinfo/talk
Re: [OSM-talk] High Cartographic Quality Label Placement on OSM-based Map
On Thursday 11 July 2013, Maxim Rylov wrote: Hi all, We are pleased to announce that a new web map based on OSM data has just been published. You can see it on http://openmapsurfer.uni-hd.de/ (OSM Roads (New) layer). One thing that greatly distinguishes it from other online maps is the quality of map lettering. Could you explain in what ways this is the case. Since different types of labels are shown in various maps direct comparison is difficult. You seem to very well avoid overlaps between labels and none the less you are able to put quite a lot of them on the map but non-label feature do not appear to play a role in label placement and there are some strange priorities. Is the algorithm available as open source? Greetings, -- Christoph Hormann http://www.imagico.de/ ___ talk mailing list talk@openstreetmap.org http://lists.openstreetmap.org/listinfo/talk