As as representative of GrubyPunkt Urban Laboratory, I participated in the workshop Dynamic Public Spaces during BMW Guggenheim Lab which was in Berlin on 27 of July 2012. As the theme was related with our own projects, we would like to share our own thoughts regarding using video analytics tools in tactical urbanism.
Before the workshop, there was the presentation given by Anthony Vanky @tvanky from Boston, one of the member of the MIT Senseable City Lab. He asked some important questions which arise nowadays around using new technologies in urbanism, processing data from multiple sensors, video data etc. Let’s make a small overview of this questions and share our own thoughts.
What we can do with that amount of the data? This question has been answered by the members of BMW Guggenheim Lab, at least in relation to post-processing tools that could filter out the data that could be of some interest. As an example, we can give the possibility of extracting human walk path from multiple video frames captured for some period.
This kind of video analytics algorithms and software is a standard in modern video surveillance systems and has been included as building block in modelling software like Matlab [1], to give an example. Also high-performance hardware implementations are available [2]. So it is expected that many developers will include such features in their software or hardware, and many equipment in the close future will be capable of doing such analysis ‘on the fly’.
The technical aspect is very interesting, fascinating in this case. However we see even more important aspects of this issue, for example the purpose of using such tools. When should this analysis be performed? How should the results be interpreted? Let’s keep in mind this issue.
The organizers of the workshop proposed us a walk to the center of Berlin: to AlexanderPlatz. We took a lot of objects like pieces of wood, chalks, jump ropes with us. We even had a small swimming pool - but with water in bottles inside! The task was simple: to do something with it on the square. All the square was being recorded with a camera. After the intervention on the square took place, we went back to the Lab and could observe the tools of transforming the images from the square, extracting useful data. We saw how people were changing their daily routes. The application showed us images identifying each person and then tracking particular people. Tracking has been performed before and during the action on the AleksanderPlatz [3]. I must say it was exciting game.
The system was quite accurate, however there were many considerations of behavior of the algorithms for the detection of people in some particular situations. For example, a situation when a couple is walking shoulder to shoulder could be difficult for an algorithm. Many technical details of the solution has been discussed.
Showing the results of analysis, another question was asked by BMW Guggenheim Lab: How many data did you produce yesterday? This is a good point for a reflexion. Realize how many data you produce each day. As we live in the world in which we try to re-use everything that we can, the usage of the produced data seems to be natural. However, as a next step we see finding the answers to the points which are very important: purpose, justification of using such complex tools, unique features of presented approach. Why do we need so much data?
We would also like to refer to another interesting issue related to the workshops. How do people create a map of the city? Most commonly used maps of the city use two-dimensional Euclidean space representation. This is the way how traditional map works, this is how Google Maps work and this is also the map on which we plot the captured tracks of people using video analysis.
Differently, humans build maps that base on points and lines strictly related with their subjectively important places of interests, memories and emotions. One of the best illustrations are the emotional maps made by Kevin Lynch, also cited by MIT Sensable City Lab. Lynch - urbanist, sociologist and also – what seems to be the most important - people-watcher was the creator of the mental maps [4]. Because each of us has own individual and subjective map which can have, in contrary with those traditional, the white spaces, can exist without any proportions and consists the ‘errors’, if we compare it to Euclidean representation.
Due to subjective point of view, our own map is orientating us in our city and what is the most important point: serves us as practical tool. So by registering the emotions as an effect occurs the mental map. Using video analytics, we can have an image with tracks of the moving people. This image without any doubt give us and translate a little those complex messages by which we are surrounded or even more, overloaded. It is good to know and have it as much transparent as we can have it. The question that comes to our mind is if we can build some translation between the subjective map and the Euclidean map. Are captured tracks of people kind of equation from which the solution of this problem will emerge?
Human beings base on subjective maps, but their behaviour can be objectively measured. And to act for improvement the situation in the city, we can both rely on the precise data and have insight into subjective perception of users of the space. That’s why I like the question How to measure to make it more understandable?. The importance of transparency of the data, the clear method of transmission in order to understand the dynamics by which we are all surrounded.
To summarize, an interesting tool has been presented, and we think its real value in improvement of the cities can be found. However, we should go with this one step further: from being impressed with the possibilities of a tools, with amount of data, to the real life scenarios of using them. It has been already shown ‘how good methodology of transforming and showing the data we have’. We generate nice plots, nice data, we can buy it and think what to do with such a pretty image… Now the question is: why, for what? When this is answered, the data will be used if there is a real reason in the real given public space where problems have been detected.
We must keep in mind that the city intervention has its constrains, basically reflected in reduced time and resources. This is fundamental to make correct decision about tools that are going to be used during an intervention.
The toolset of urban laboratory cannot be selected basing on fascination - it must be proven that in some given subset of urban problems video analytics, or any other innovative tool, does its job better, more efficient than others.
For this, we in GrubyPunkt Urban Laboratory put much importance to some aspects of the experiment which has not been so exposed. We put first the real need of a public space. What is the need and how we can respond to this need using features of the tool.
An answer to this question has been also signalized in the BMW Guggenheim Lab’s experiment as the selected space was not chosen without reflection. It was one angle of the super vivid square with smallest quantity of the passer-bies. So the aim was to make this angle more alive. In fact, this aim was achieved and the result has been measured, which is an advantage of the presented system. The interaction changed the behaviour.
The question: Can we change the pattern? has been answered. We can, and we also can measure this change. However, why it has been performed? Possibly the experiment could be modified to be more complete. Using the same tools, some real change, even small one, could be introduced. By real change, we mean providing persistent value to the inhabitants. Sometimes this real change can be slight and boil down to few recommendations, but can bring a real positive effect. We are looking for this kind of outcome in our work.
We think that now it’s time to show how this change and this analysis can be converted into real positive transformation in the city. Because there are still a lot of question to set and problems to investigate and resolve. We have bricks that we need, but now we must build a solution for the city. Because obviously, the aim is not to follow, track, change behaviour, change pattern, nothing like that. And we must remember it. The aim is to improve the situation in the city.
We feel there is a lot to do in this area, as technology offers us new tools that we can use wisely, combine with traditional approach. People tracking is one of the tools and has been present for some years, first in the surveillance and military market, and now it’s about to find its use in urbanism. But there some more techniques that are so new that now it’s even hard to think how they could be applied in our scenarios. As en example, self-learning neural network based video analytics systems [5], face emotion recognition system, gender recognition, age recognition [6] etc.
There are many fascinating tools and many will come in the next decade. What we in GrubyPunkt Urban Laboratory see most important is go beyond this fascination phase and find and show real uses. Maybe starting with very basic, small observation, reflection and then usage of the data, analysis that lead to some simple but useful improvement of our nearest public spaces. If we find this uses, we will prove which tools are really useful for local communities.
[1] Computer Vision System Toolbox. Matlab, Mathworks - http://www.mathworks.com
[2] Multi-core Video Analytics, Eutecus Engine http://www.eutecus.com
[3] Video footage of the Dynamic Public Spaces Workshop http://www.youtube.com/watch?v=vec_rdp2yB0
[4] fragments of “Image of the City”, a book by Kevin Lynch http://books.google.es/books?id=_phRPWsSpAgC&pg=PA46&hl=pl&source=gbs_toc_r&cad=3#v=onepage&q&f=false
[5] AiSight, BRS Labs, http://www.brslabs.com/
[6] Centro Technolóxico Gradiant, http://www.gradiant.org/en/research-lines/human-sensing.html
![As as representative of GrubyPunkt Urban Laboratory, I participated in the workshop Dynamic Public Spaces during BMW Guggenheim Lab which was in Berlin on 27 of July 2012. As the theme was related with our own projects, we would like to share our own thoughts regarding using video analytics tools in tactical urbanism.
Before the workshop, there was the presentation given by Anthony Vanky @tvanky from Boston, one of the member of the MIT Senseable City Lab. He asked some important questions which arise nowadays around using new technologies in urbanism, processing data from multiple sensors, video data etc. Let’s make a small overview of this questions and share our own thoughts.
What we can do with that amount of the data? This question has been answered by the members of BMW Guggenheim Lab, at least in relation to post-processing tools that could filter out the data that could be of some interest. As an example, we can give the possibility of extracting human walk path from multiple video frames captured for some period.
This kind of video analytics algorithms and software is a standard in modern video surveillance systems and has been included as building block in modelling software like Matlab [1], to give an example. Also high-performance hardware implementations are available [2]. So it is expected that many developers will include such features in their software or hardware, and many equipment in the close future will be capable of doing such analysis ‘on the fly’.
The technical aspect is very interesting, fascinating in this case. However we see even more important aspects of this issue, for example the purpose of using such tools. When should this analysis be performed? How should the results be interpreted? Let’s keep in mind this issue.
The organizers of the workshop proposed us a walk to the center of Berlin: to AlexanderPlatz. We took a lot of objects like pieces of wood, chalks, jump ropes with us. We even had a small swimming pool - but with water in bottles inside! The task was simple: to do something with it on the square. All the square was being recorded with a camera. After the intervention on the square took place, we went back to the Lab and could observe the tools of transforming the images from the square, extracting useful data. We saw how people were changing their daily routes. The application showed us images identifying each person and then tracking particular people. Tracking has been performed before and during the action on the AleksanderPlatz [3]. I must say it was exciting game.
The system was quite accurate, however there were many considerations of behavior of the algorithms for the detection of people in some particular situations. For example, a situation when a couple is walking shoulder to shoulder could be difficult for an algorithm. Many technical details of the solution has been discussed.
Showing the results of analysis, another question was asked by BMW Guggenheim Lab: How many data did you produce yesterday? This is a good point for a reflexion. Realize how many data you produce each day. As we live in the world in which we try to re-use everything that we can, the usage of the produced data seems to be natural. However, as a next step we see finding the answers to the points which are very important: purpose, justification of using such complex tools, unique features of presented approach. Why do we need so much data?
We would also like to refer to another interesting issue related to the workshops. How do people create a map of the city? Most commonly used maps of the city use two-dimensional Euclidean space representation. This is the way how traditional map works, this is how Google Maps work and this is also the map on which we plot the captured tracks of people using video analysis.
Differently, humans build maps that base on points and lines strictly related with their subjectively important places of interests, memories and emotions. One of the best illustrations are the emotional maps made by Kevin Lynch, also cited by MIT Sensable City Lab. Lynch - urbanist, sociologist and also – what seems to be the most important - people-watcher was the creator of the mental maps [4]. Because each of us has own individual and subjective map which can have, in contrary with those traditional, the white spaces, can exist without any proportions and consists the ‘errors’, if we compare it to Euclidean representation.
Due to subjective point of view, our own map is orientating us in our city and what is the most important point: serves us as practical tool. So by registering the emotions as an effect occurs the mental map. Using video analytics, we can have an image with tracks of the moving people. This image without any doubt give us and translate a little those complex messages by which we are surrounded or even more, overloaded. It is good to know and have it as much transparent as we can have it. The question that comes to our mind is if we can build some translation between the subjective map and the Euclidean map. Are captured tracks of people kind of equation from which the solution of this problem will emerge?
Human beings base on subjective maps, but their behaviour can be objectively measured. And to act for improvement the situation in the city, we can both rely on the precise data and have insight into subjective perception of users of the space. That’s why I like the question How to measure to make it more understandable?. The importance of transparency of the data, the clear method of transmission in order to understand the dynamics by which we are all surrounded.
To summarize, an interesting tool has been presented, and we think its real value in improvement of the cities can be found. However, we should go with this one step further: from being impressed with the possibilities of a tools, with amount of data, to the real life scenarios of using them. It has been already shown ‘how good methodology of transforming and showing the data we have’. We generate nice plots, nice data, we can buy it and think what to do with such a pretty image… Now the question is: why, for what? When this is answered, the data will be used if there is a real reason in the real given public space where problems have been detected.
We must keep in mind that the city intervention has its constrains, basically reflected in reduced time and resources. This is fundamental to make correct decision about tools that are going to be used during an intervention.
The toolset of urban laboratory cannot be selected basing on fascination - it must be proven that in some given subset of urban problems video analytics, or any other innovative tool, does its job better, more efficient than others.
For this, we in GrubyPunkt Urban Laboratory put much importance to some aspects of the experiment which has not been so exposed. We put first the real need of a public space. What is the need and how we can respond to this need using features of the tool.
An answer to this question has been also signalized in the BMW Guggenheim Lab’s experiment as the selected space was not chosen without reflection. It was one angle of the super vivid square with smallest quantity of the passer-bies. So the aim was to make this angle more alive. In fact, this aim was achieved and the result has been measured, which is an advantage of the presented system. The interaction changed the behaviour.
The question: Can we change the pattern? has been answered. We can, and we also can measure this change. However, why it has been performed? Possibly the experiment could be modified to be more complete. Using the same tools, some real change, even small one, could be introduced. By real change, we mean providing persistent value to the inhabitants. Sometimes this real change can be slight and boil down to few recommendations, but can bring a real positive effect. We are looking for this kind of outcome in our work.
We think that now it’s time to show how this change and this analysis can be converted into real positive transformation in the city. Because there are still a lot of question to set and problems to investigate and resolve. We have bricks that we need, but now we must build a solution for the city. Because obviously, the aim is not to follow, track, change behaviour, change pattern, nothing like that. And we must remember it. The aim is to improve the situation in the city.
We feel there is a lot to do in this area, as technology offers us new tools that we can use wisely, combine with traditional approach. People tracking is one of the tools and has been present for some years, first in the surveillance and military market, and now it’s about to find its use in urbanism. But there some more techniques that are so new that now it’s even hard to think how they could be applied in our scenarios. As en example, self-learning neural network based video analytics systems [5], face emotion recognition system, gender recognition, age recognition [6] etc.
There are many fascinating tools and many will come in the next decade. What we in GrubyPunkt Urban Laboratory see most important is go beyond this fascination phase and find and show real uses. Maybe starting with very basic, small observation, reflection and then usage of the data, analysis that lead to some simple but useful improvement of our nearest public spaces. If we find this uses, we will prove which tools are really useful for local communities.
[1] Computer Vision System Toolbox. Matlab, Mathworks - http://www.mathworks.com
[2] Multi-core Video Analytics, Eutecus Engine http://www.eutecus.com
[3] Video footage of the Dynamic Public Spaces Workshop http://www.youtube.com/watch?v=vec_rdp2yB0
[4] fragments of “Image of the City”, a book by Kevin Lynch http://books.google.es/books?id=_phRPWsSpAgC&pg=PA46&hl=pl&source=gbs_toc_r&cad=3#v=onepage&q&f=false
[5] AiSight, BRS Labs, http://www.brslabs.com/
[6] Centro Technolóxico Gradiant, http://www.gradiant.org/en/research-lines/human-sensing.html](http://25.media.tumblr.com/tumblr_m92oxhpMK31rvv2f3o1_500.jpg)
