0:00:15and the ladies and gentleman and it's my blessed to be here in sandy uh
0:00:22but no and in this uh conference and then i uh i have a real
0:00:29my presentation uh
0:00:32i had quite practical and at the channel level i'm an engineer i'm not the
0:00:38mathematician so hopefully this is okay to you
0:00:43so um
0:00:45and i will talk about the mess envision solutions and especially if although forest the
0:00:51industry
0:00:53and uh maybe the next slide that was more about motivation this is accomplished every
0:00:59of my university in that in that a lot better and the and this area
0:01:05is that so called in is like this trick area so it's full of the
0:01:09lakes and forests
0:01:12and the forests it is a big business big in the study in that area
0:01:18so actually it's the area where there is the most then set at a number
0:01:24of while males and solving this in the world
0:01:28so it's quite natural uh since the six kilometres from our university there is a
0:01:35big by a male and there are two other paper mills quite close
0:01:39um about um
0:01:42L one university uh it's one of the three a technical universities in finland and
0:01:49here are our main focus areas
0:01:53so green energy and technology
0:01:58sustainable value creation and international have of R S and relations and you can see
0:02:05because we are at the border is the closest to university to rest of the
0:02:10so that's quite a match
0:02:14so um and also my uh research area in must envision connected the forest industry
0:02:21it's quite close to these calls
0:02:24as you can later like to see
0:02:33so this is our motivation this is probably very clear to you um what is
0:02:40the motivation of computer vision and machine vision but from the practical point of view
0:02:45at six i especially these and these are not of fake pictures these are real
0:02:49pictures this is a person in uh mail paper mills laboratory to in serious uh
0:02:57quality assurance testing
0:03:01so it's mostly damman wiley and you as you can see it really minor
0:03:06so there are some are based but it's
0:03:10it's uh bring it on a piece of paper board and then this person is
0:03:15used in magnifying glass checking
0:03:19is it is it okay is the quality okay what this the prediction of the
0:03:24usability of this um at the
0:03:27and that this is how it should be done
0:03:30so bill built in uh matching wits that's the same thing and the in the
0:03:37standard last way one of the channel and C is there is that this depends
0:03:42very much on human body and the uh and uh be right is based on
0:03:49individuals a preference is still there is a slight variation
0:03:54they are and also this monday morning and uh friday afternoon effect too
0:03:59and this is a standardised way to do it
0:04:03since this information goes from paper mail yeah males from this uh paper manufacturers to
0:04:11a printing house that what is the quality of the product is any
0:04:16and actually printing houses are doing their own similar tests so as to where E
0:04:21Y
0:04:23that opinion these the same
0:04:26so this uh would be a standardised solution is
0:04:32so uh
0:04:36another thing of course is that this is similar don't knows task and some that
0:04:40cell or spencer's at in the industrial environment
0:04:46and uh this motivations i think you have a scene that it's fast the more
0:04:51accurate less expensive i wouldn't say gee but i would say less expensive because that
0:04:57must be sent to quality and easy to use is important
0:05:02this
0:05:03um people who will be systems they sometimes will get that it must be easy
0:05:08to use
0:05:10and sometimes the realm us to do not some in for my mation or two
0:05:15months options which you never actually
0:05:18so is it the use
0:05:21and it's a good thing
0:05:23and motley question of course is the replace human recent totally in this case that
0:05:29somebody ways
0:05:30so no human of we need it anymore at all a mess in instead
0:05:36ses human-based and so sometimes to keep some hint this is mainly for example in
0:05:43medical image processing
0:05:46and then perform that's the time possible to human vision system you can imagine some
0:05:51online uh solutions where you simply don't have ability to be so fast
0:05:57when a paperweight is running the that the meters per second
0:06:02so you that you can't handle that time sequence in your mind
0:06:07or simply we don't have always in system can see spectral information for some
0:06:14so this is our um
0:06:17uh
0:06:19some kind of uh starting point
0:06:23and uh
0:06:25this is just same shortly about the low-level auditory machine vision and but the recognition
0:06:31of auditory we have we have around the when the people
0:06:35doing this results and we have a it you kate the twenty seven uh doctors
0:06:41in this area so in machine vision and many times the uh these theses they
0:06:48are connected to real problems it's comes from industry go from bases
0:06:56and we have around one million euros but here is how much it and six
0:07:00hundred thousand euros just
0:07:02is um
0:07:05um external funding and of course L one bit main goal is to produce a
0:07:12speed of common one is where these ideas could be implemented as soon as possible
0:07:17in practice
0:07:24um generally however is not only supporting building a built in a visual inspection systems
0:07:31like here we have a many other areas like a fire model correlation here is
0:07:38a bacterial a matter
0:07:40so that sensors have a it's a consist of that there is not on any
0:07:46electronics and then intelligent robotics and object recognition here face recognition and one area which
0:07:54we especialised this medical images analyses and especially analyze civility not images so to see
0:08:01some uh this easy some problems for example cost by diabetes
0:08:07so that is also area
0:08:13um if we think about the visual inspection actually it's uh um
0:08:20it's a it's a actually what white lots color correction so there are many other
0:08:27universities involved and also many real eight comp running so a lot of a lot
0:08:33of all operation is needed and if we need to think about how one of
0:08:39our main objectives is the develop novel results efficient and environmentally sound image in image
0:08:47processing and image and the lies in met thoughts for machine vision applications especially for
0:08:53overall quality assessment and control so the idea is to all these single measurements together
0:09:01to be able to control in general
0:09:06the whole industrial process
0:09:10and we aim to offline and online solutions obviously that is a lot part of
0:09:17our example in the preview on the previous slide that it is offline
0:09:23uh solution because it happens in the laboratory but then we need also online solutions
0:09:29directly uh in the industry
0:09:34and mainly we are focused on forest and printing industry
0:09:39so at the lower product already and the factory
0:09:42about thirty level
0:09:48and um if we think uh how
0:09:53what is actually happening there you see that this contains look over different the industrial
0:10:01is text there so it's actually quite complex uh combination before you end up to
0:10:07get to have a lot of paper there are so many things happened before that
0:10:13so in bright this we have a some would might be real then we have
0:10:19well and then paper and finally some print so something you can print or
0:10:27uh and uh if you think wow why to do it okay the results efficient
0:10:34uh resorts in efficient uh environmentally sound production with no one data using less rollment
0:10:41the real water and energy and actually this is a cool thing because this to
0:10:47be environment three uh one shows uh use the same as the road using it
0:10:54efficiently so there is no contradiction between these things because if you use less raw
0:11:01material water and energy you save money and you also save environment
0:11:07and the no other known quality means that the uh you use should put quality
0:11:13as low as possible and it still feasible quality so it's still suitable quality you
0:11:20can always book the parameters so high that the quality is high but you waste
0:11:28a raw material what the energy and so on and you waste money
0:11:32so as low as possible but still reasonable quality that's the idea so that these
0:11:38than on quality and we'd without having some measurements and they interpretations we'd be gone
0:11:44we don't know what these what these that quality so that these what we try
0:11:50to write the so
0:11:53actually our uh meals in finland they uh environmentally in very good condition because the
0:12:01lakes around a white lee
0:12:05and that we actually condo for two only you of lakes because then the for
0:12:10its we like to do so we must really think of environmentally seems to
0:12:19so um
0:12:21idea simple in the way imagine they the automatic analysis and the in the forest
0:12:29industry
0:12:32uh we talk about this will also line quality has assessment business but this done
0:12:37right traditionally manually
0:12:40and the campy time calm you consuming and labourers to
0:12:46then we have online quality assessment and then tool for real time control actually these
0:12:52two areas and the research done their support this idea
0:12:59to have a some kind of real two-dimensional uh quality control based and to advise
0:13:05dimensional units
0:13:08and uh what we have a
0:13:12study we have had we see this as a as a collection of uh research
0:13:18approach it so we invested
0:13:21so we have one uh rotate which is dealing with power
0:13:26so how to put this uh raw material in the bases for paypal make
0:13:33and what could be done there
0:13:36then the one thing is that what can we do um
0:13:42connected to a paper messy so when real paper making is happening so and there
0:13:47is a long paper my signal for one hundred be there with the long volume
0:13:52along more what we can do there
0:13:55and of course these are connected so if you know about the quality of raw
0:13:59material here you can you can put this information here and you can take advantage
0:14:04of that
0:14:05and then this is your final product okay um here you can't do anything anymore
0:14:13it's already they are so paper so after that you can only predict how to
0:14:17print on and what is the pretty that uh applicability and run ability when doing
0:14:23some pretty
0:14:25and then
0:14:27then uh we test the quality of the paper and actually finally interesting thing is
0:14:33that how we see that you meets or something pretty become paper that can be
0:14:40mortal human beings set preferences is connected to some physical much innovation measurements
0:14:48so that's is that and the system will be nice that you put the print
0:14:53into that find a much emission system any else that how human beings are uh
0:15:00but if you're in this but it's
0:15:03"'cause" uh the quality of an image but in it of course it's very important
0:15:08you don't want to see some bad quality images but what is the bad quality
0:15:12and what is that what did when i ask if i thought make the test
0:15:17we you i assume there are some but em sis but there are some form
0:15:22of that are still so this is a big question and this actually relates to
0:15:25any image information problem whether it's print date or whether it's speech it's exactly the
0:15:32same problem
0:15:34problem
0:15:36it's very uh and fess in a be much fascinating problem actually and what typical
0:15:45i will solve some results about this before and the steps here
0:15:54and what one wants to remember this that uh need for these kind of research
0:16:01must come from the industry so every single project we have a start date has
0:16:06started in that way that we have once industry what do you need and then
0:16:12we have stopped at the check what we can
0:16:15so it's actually it's not uh selling your ideas it's uh
0:16:21asking needs and uh formulating the questions
0:16:33okay us then i would like to talk about some um
0:16:38things uh which we must consider and one is to recognize imposes phenomena
0:16:45so actually what we should and this that is the question and the who knows
0:16:53the ground rules and why and how
0:16:56and
0:16:58yeah and it is it really possible the model the expert knowledge
0:17:04and uh this is a good question because we must we must talk to be
0:17:10a those experts there is that the what they would like to know and what
0:17:17we could see and uh
0:17:20sometimes it's islands in if the data that is very mobile so even the expert
0:17:25have an ever seen they the from some certain industrial processes
0:17:30so this really needs need something in and actually i have an all this when
0:17:35we are having them some uh comparisons that the and i is that where is
0:17:40the ground rules if i'd use your data and uh and some might be just
0:17:46looks good but that's not you know
0:17:49you should really be fine ground and maybe to other researchers publish your mate complex
0:17:55operation ground
0:17:57of course in this kind of uh industrial pros project said this comes a little
0:18:02bit be late because firstly analyze the possibilities to make a real part that are
0:18:08based on that it is it's but it will come anyway
0:18:13and is this possible that's and that's and other is you sometimes it might be
0:18:19very seen that the sum intensity values that everything but
0:18:27you don't know before you think it's
0:18:30mister in optimality
0:18:33so what utterances in image and so uh there might be some illumination problems or
0:18:41it might be a typical place in there or you need some uh not reflection
0:18:48information but that some transparent information
0:18:52and then we call to this uh do we need multimodal information that many times
0:18:57we need so we have a transmission in its we have reflection immediately may even
0:19:04have some extreme images
0:19:08we call the different scales there and how to put this together so we have
0:19:13multimodal information and how to how to put it together and whether it supports to
0:19:18each other
0:19:22then understand in measurements
0:19:25so a bible college on unclear categorization some more clustering need it so that we
0:19:33know enough about they may not then we have the first define what are the
0:19:37classes there
0:19:39so clear when it's be about the like categorization then it's just classification but many
0:19:43times as a question of this especially when it's number of data that what are
0:19:48the other clusters
0:19:51so this might be a challenge to all of course accuracy and computation time if
0:19:56you think of this uh paper with running that the meters per second you have
0:20:02two options you where infrequently uh to sampling but the image resolution is uh low
0:20:11or you have a high resolution but then you don't do it so frequently
0:20:17and our idea is to go to this overall quality management and approaches process control
0:20:24so step by step we more accurately know which kind of uh paragraph we will
0:20:30get
0:20:32and this is this is about how been and actually this goes into those of
0:20:37previous steps as
0:20:39as i mentioned
0:20:41and there are some certain uh process phenomenon so probably means that the uh we
0:20:49make a raw material for people making
0:20:52so we need to check some certain things like uh i other some extra particles
0:20:57they are which are not wanting to be there and the basis actually about suspense
0:21:05and he meets so it's a liquid these is the right this is the right
0:21:09about C have been some i'm one D the objects
0:21:14and uh number of a number of this and what they do or distribution of
0:21:20this on one be the object of like say that well
0:21:25so mess are in these things
0:21:27different kind of uh i things even for a sense and spectral image and then
0:21:33this analysing part
0:21:37so here are some examples here and one thing you know very important thing of
0:21:42course is to understand the uh fibres that fibres uh distribution and position
0:21:50so i'm not will industrial data for the uh
0:21:56by new image and solutions and uh i'll idea is to pull inside the pipelines
0:22:03which is uh have not been done so our data is totally novel so wipers
0:22:09are you know interest the particles buckles and sales these bubbles the lapel approaches there
0:22:15is some you know their problems there
0:22:18and then how the form and the model the ground rules and this need something
0:22:22can because if it's a mobile make the so there is no a human expert
0:22:25that's believable expect is about
0:22:28but the there is six but is about the uh i mean
0:22:34so that can be used
0:22:40uh and uh then uh one thing there is a segmentation the fibres another particles
0:22:46in suspense and emits is still this is really weak
0:22:49lee we thing and we have a done that the uh results and why this
0:22:57piper's must be uh and the last is that you should know the uh three
0:23:03species there it may sound strange that don't they know what is there but it's
0:23:08not necessarily always the same tree and same kind of three material so uh the
0:23:14know about these three C Cs is important and again this means that the this
0:23:19viper is broken in the back leeway that later when you breed
0:23:24how can they need that kind of uh fibres it's bad property
0:23:29so the to detect number of things
0:23:32is important
0:23:37and we have some applications about this
0:23:41if you call to my list of uh applications you see that should have that
0:23:46six
0:23:48then another use you uh not only five course but problem side important because particles
0:23:55tell about some approach the situation and how they daily the daily and in that
0:24:00way that um when there is a certain amount of purple bubbles the process is
0:24:06uh okay so that such a certain size and certain volume distribution and how they
0:24:13do it now when they are aligned in a factory they do it in that
0:24:17way that they'll and the process for let's say two hours to be sure that
0:24:21everything is okay but one hour would be enough
0:24:26but you don't know that if you don't have an image
0:24:30nonetheless
0:24:31so you put even got terms is processed by means of halls and that would
0:24:36be you say in
0:24:38but you need in leeds analyze the solution for that so we have real the
0:24:43field of that kind of a solution and we will shortly present this is very
0:24:49impressive results and icp our conference in japan in november this is a couple operation
0:24:55results with the tech technical university
0:24:58we propose that matters
0:25:00and actually this is a this is a quite challenging problem although that looks so
0:25:05originally quite easy and if you have a some uh proposals or something to the
0:25:13to this problem i'm very happy here you or your I S
0:25:20so it is not easy
0:25:23since it because these problems are not that we're a circular problems so they may
0:25:30be deformed so actually we are dealing with the P Cs or for some kind
0:25:36of circular say and we need quite the white mats thinking how to make it
0:25:42that efficiently and then we must detect the problems in order to be able to
0:25:48estimate what so from problems we estimate the volume has some value there and sof
0:25:54detection of problems as important than in this case of course detection of the become
0:25:59problems is more important than the than the small ones
0:26:03but also this distribution map cell from distribution experts can tell something
0:26:11and uh we have actually ended up matt's of thing you know feel that in
0:26:16how we get some kind of suitable it speaks
0:26:21and enough of those and how we start to combine of these pixels because then
0:26:27never ideal unfortunately it's a liquid and it's also problem of three-dimensional imaging
0:26:36there so but uh i must say that then we have got quite a very
0:26:41good results and surprisingly this uh this uh false negatives some false positives they start
0:26:50to compensate each other and uh not most of that we can detect right so
0:26:57we can really get true positives but the so surprisingly um well then of course
0:27:03the question these that what is the what is the ground and that because of
0:27:08bases you we have given the experts to select a level of confidence
0:27:16the whether there is a war or and solar or totally unsure
0:27:21because it's not so clear what his answer to so
0:27:27so these contains all the nearly all the possible problems i enumerate the in the
0:27:32couple of slides ago
0:27:35okay but that the that the say that the pc is challenging from the uh
0:27:39point of view we mix process
0:27:43some people sometimes say that all you test an application itself see but we are
0:27:49dealing with the fundamental issues of image processing
0:27:56okay then and that other thing that the uh okay fibres
0:28:02um couples
0:28:04so five resigned the liquid
0:28:07and this is bob suspension bubbles in about suspension but this is the right biopsy
0:28:14so when the power was ready each tried to be more to uh to a
0:28:19paper mail and this is really a tried C and E there should be only
0:28:24uh suitable rollment the we only but all these so with this dog you the
0:28:29there's something and one it's um the particles because of what's something's there might be
0:28:35even this and they are
0:28:38so any anything
0:28:40uh
0:28:42the right of course the different kind of categorization even plastic
0:28:47or some unwanted what would my
0:28:51so it would be good to categorise these
0:28:54different kind of uh the particles "'cause" that's there was the quality of the biopsy
0:29:00if there is no the particles it's a it's a great but the problem she
0:29:05for paper make
0:29:10and so we need to a here the segmentation and then classification of these this
0:29:15that the particle physics
0:29:18once again we are dealing with the quantum fundamental issues of image processing and analysis
0:29:27and here is a slight problem because we must do this testing that the laboratory
0:29:32label
0:29:34see scenes uh
0:29:36uh if we do it online
0:29:40then uh we wouldn't have time to do analyses so what we have done we
0:29:45have a uh at the laboratory we have a creative israel about six and in
0:29:50that way that the experts have really put the ground rules there
0:29:55and then we can test whether we can recognise this separate that the particles
0:30:02at the moment there is no analyses
0:30:05you just put up sixteen and you hope that you will get would like
0:30:10so that's motivation to build the system to do that and S
0:30:14okay then let's go to the know now we worry at in the powerpoint so
0:30:19a would deinterleaving liquid and do some processing there it can be column mechanical then
0:30:24you get that right biopsies then you put these power seats uh to a paper
0:30:29machine and then you start to make paper
0:30:33and the um main question is that it is that it presents and this represents
0:30:39a paper a bit so that what is running from the beginning to this and
0:30:43when the product is ready uh the question is that can we see here or
0:30:47something
0:30:49something which we should uh is that the react to something and actually um now
0:30:55there are some line scan or switch to body weekly scanning but that it's of
0:31:00course different than the two-dimensional to time instant no uh real time image and uh
0:31:06because of scanning
0:31:08we actually can on see this kind of till the problems which you may see
0:31:16basically simple maybe you somehow can see that and uh of course of human being
0:31:21can see that's that the meters per second so close to plastic on C so
0:31:26you're we must rely on these mis ordering systems and line scan risk on to
0:31:31it because the coast line by line so idea is to take a real the
0:31:36dual diamond two dimensional units uh challenge there is that we need eight income it
0:31:42is
0:31:43and put all these images together
0:31:46and so it's quite the real-time problem and also build in an image problem too
0:31:53because of uh some uh geometric uh imaging problems and things like that but that
0:32:01we can do so we can find
0:32:05fine these kind of because these problems these sideways problems they are because of some
0:32:13uh not so accurate tuning at the beginning of the process and you can tune
0:32:18the parameters there but you must know in which way so this gives feedback
0:32:24they're to do this do you
0:32:29so how to produce and use real-time two dimensional web white image state the problem
0:32:35of or on online uh control
0:32:41um then we can go uh further in our analysis uh we can we can
0:32:48focus on at some uh more accurate level uh and uh get some information and
0:32:56why we uh this is some kind of closer close up a information from that
0:33:01with uh paper web that area
0:33:06and there are some uh phenomena we should see that how actually the white borders
0:33:12are grouped there that's important but if that my signal then we can then we
0:33:19will get nice paper and print on it then we apply but there might be
0:33:22some grouping problems and this that was also about
0:33:27do
0:33:29and uh here we can do this a multimodal problem because we have our lower
0:33:34low resolution high resolution images we have a reflectance it means it means we have
0:33:40transmit dance emits we have seven of images and how to help able to put
0:33:44these things together and we can have some other kind of measurements to so we
0:33:48have a bit different kind of a units in many thoughts and how we can
0:33:55put them together so here for example the segmentation and analyze of three five or
0:34:01flocks using transmis tense images
0:34:05and when the orientations are nice then it's okay maybe when you have a at
0:34:10the right that the other paper you know this that it that the there is
0:34:14very nicely there some certain direction about another direction that some and the that's because
0:34:19of
0:34:21on purpose it's made in that way
0:34:27and uh
0:34:29so detection of surveys orientation is so one thing to uh check and then of
0:34:36course the challenges here that how we get these correspondences when we have a different
0:34:41kind of image state that from the same
0:34:46so we have a correspondence problem they are
0:34:49so how to do image registration
0:34:58and why we are interested in surface orientation is that it's affects directly but
0:35:07and then after we have got this paper then we can test what we call
0:35:13and we have build the system to do this that and the vices
0:35:20and so is to do this
0:35:23the similar kind of solution to everybody
0:35:28not depending on the human brain
0:35:31and let's see what we can see from that so we can we can have
0:35:35a different find of uh measuring devices here and the what we i'm gonna speech
0:35:41in this uh for one is an to be in and evilness of print in
0:35:45misery you know name it so it's modeling so we should pretty that the uh
0:35:51if we bring the blue or lost by their how well how you nibble form
0:35:56this
0:35:58and can have some kind of my sermon some estimate and it's called model in
0:36:03the in the printing industry
0:36:06"'cause" if that was crying is not uniform it disturbs you i and you think
0:36:11it's about three i think this is maybe
0:36:14but pretty but you can see if you if it wasn't
0:36:21so that's a one thing somewhat yeah modeling detect the modeling
0:36:25another thing is that the
0:36:28how will this step putting in happens on the survey so it's called haley a
0:36:34test so detection of missing dots in that this but it is bought to say
0:36:39about this that this is mostly about some uh frequency feel that in of different
0:36:44kind of a yeah yes so that's a technique and then the question is then
0:36:48how we weight of those different frequencies
0:36:52it's as the linear combination or are there some coefficients with the with the combined
0:36:57or something like that here this is very simple test the new print of on
0:37:02a paper sleep uh don't red dots and then you simply calculate which starts are
0:37:08missing because it means that the so face is not so cool and this simple
0:37:12man in this woman in the picture was really count in and where is that
0:37:17when the at the missing dot from the beginning and he may shirt she mesh
0:37:22shirt that in millimetres and role there that ninety seven millimetres from the peak that's
0:37:27the quality while the longer the better
0:37:32so and we also made it is uh there's the one we used it is
0:37:36there's another and we automated the this column
0:37:41so ago that was automatically this of course the question there's a one challenge that
0:37:47what is the mission of how much must be missing or and then person be
0:37:53left there "'cause" of course this is natural situation this is not by nist
0:38:01then uh um some small be there is how they sexy carissa must of these
0:38:08is about that how these uh
0:38:12this actual but it looks
0:38:16what we can see from that from a single small not
0:38:22it's obvious where about the predictability this is a part of an ability in printing
0:38:27the it's called the peeking test that you press harder and harder that the that
0:38:33uh test sleep until it breaks
0:38:39or some i'm a other that meets this will start to happen to
0:38:43so for usually uh and this is the border it's they're starting to see some
0:38:49phenomenon the subways and when it's a paper it's imprecise the der of
0:38:56so it totally breaks and the question is that how long you grow until it
0:39:01starts to add to your break and once again people are doing by and calculating
0:39:08how many millimetres from the beginning to that place and of course it's a at
0:39:12the request and we have this happens this a one D uh phenomenon of the
0:39:18early
0:39:20and we have a thesis about
0:39:24this topic to so this terms of butter unappealing the longer the better
0:39:28from the printed point of view
0:39:34and uh here is an example about the speaking tests he actually here you see
0:39:39what happens the paper it simply start to tear off and this is not what
0:39:46if the value and what and or it usually neighbour will S is that the
0:39:51overall but there will become sums of place the effects there and this is able
0:39:57to a little bit harder to define because there is no except order here is
0:40:02quite clear that simply breaks here and that's the distance
0:40:08and actually this company's we sell in this device it's called laplacian technologies our doctors
0:40:14have a used up list a all pole one and they yeah selling this product
0:40:20uh for this text
0:40:26and this is about this missing don't the problem that
0:40:31this print starts here and you put in the pixels and then you start to
0:40:36check that how many pixels up the uh missing and that when the L one
0:40:41is the
0:40:43is the definition i is that where these come from and i think in a
0:40:47laboratory level in those paper means they have just decide that it's twenty
0:40:55so it's some kind of practical
0:40:57practical value but you should always have some but i think the uh this twenties
0:41:02in that way that you would never go to the end
0:41:06it's comes from that it's better not to say that like one of our doctoral
0:41:11candidate set in the doctoral dissertation examination that why is this threshold speech P and
0:41:17he replied that that's a good number
0:41:19but i don't recommend this was there will prefer that with questions after that
0:41:26okay then uh and next issue when it's really pretty print it on um how
0:41:32we can uh estimate the quality so which one which looks better and why because
0:41:39you may have a different print the brain there's different papers different images how do
0:41:46you can say that which one is better
0:41:49so what we need to we uh selected some S in recent measurements
0:41:56and we have the department of psychology from helsinki university to do in experiments with
0:42:03humans asking their opinions and then you know how to do it we don't know
0:42:07as engenders but they know how to do it so they collect be it's a
0:42:11human being subjective but uh
0:42:15what if the attributes and then to be put them together to get some kind
0:42:20of well overall image quality index and with this index we can we can have
0:42:25that
0:42:27double for example then images and we completely in order and hope that it goes
0:42:31in the same way as human being would have prefect
0:42:36this is quite challenging results research to
0:42:42uh in this results we have a couple of things to do is of course
0:42:46because we need to expect around rule so we need to digitise image so we
0:42:51must the register very uh accurately this data test images units and that's a one
0:42:57uh thing is this uh and uh occur missile inequality and uh locate in those
0:43:04areas which we have miss that one doctoral thesis is about that
0:43:11so the predicting human humans experience of image quality for print the image
0:43:20and the then we calculate the this uh like a bayesian network model that
0:43:26weights the all these simple things uh seem not seem well in a way simple
0:43:31measurements affect the most and um it's maybe not surprising that close is one of
0:43:37those modeling is very important and colour but this time it is a really experimentally
0:43:44proven because we really fast
0:43:47is a human being set a group to tell they opinions and these opinions were
0:43:52really model and compared to
0:43:55real but uh physical my matching pursuit measurements
0:43:59so this is not the case this is really the result
0:44:05and in this thesis by almost and allow you can be more
0:44:13um then uh i'm approaching the and only a couple of slides left here an
0:44:19actual level with you with numbers because these are selected slides not fifty so far
0:44:27so um and that next question is that what i've mentioned is at that the
0:44:32how well does this depend on which kind of image you
0:44:37and yes it does very much
0:44:41so then we would need the system to finalise what is the content of an
0:44:47image
0:44:49to have a
0:44:51E meets content based this one lady
0:44:55so it would change based on image content because many print this or see it
0:45:01on screen it's it has different preferences we deal than scene this and this close
0:45:08to very current hot topic uh visual object categorization so how to and especially in
0:45:15this unsupervised manner because we are not interested what are the objects at all we
0:45:20are just interested what objects are important
0:45:23and uh
0:45:26we don't even without knowing what though they are
0:45:29well let them to affect this decision making
0:45:33so we are dealing with unsupervised the visual object that the crisis and then we
0:45:39need the ground rules so we did i tracking
0:45:43and to see what are the salient features for human brains beings and one in
0:45:51a thing is that what are these images how what is the perfect test images
0:45:57that that's a that's a big question because uh you may have a some kind
0:46:03of side effects if you don't the
0:46:06landay's in which is carefully enough and that's why we have from all the university's
0:46:10media technology specialists who were preparing these images for us
0:46:17and uh
0:46:20uh
0:46:21and then uh we ended up really the to do this uh visual object categorization
0:46:27results and um well idea is uh use in self organizing map uh of for
0:46:35a clustering of these objects and finally get these categories so but the visual codebook
0:46:43use in
0:46:44self organizing map the component it and the also some uh evaluation techniques then we
0:46:52have a new uh we use uh published new images uh it said so randomized
0:46:59object so we make the tricks with the for foreground some background images and also
0:47:05saliency detection
0:47:08and this was connected to the sidetracking the results
0:47:12and all also
0:47:15also better techniques for what this categorization
0:47:19and uh we have also published the uh totally new limits that it's dealing with
0:47:27up static images where we are interested this ci tracking when we have a collapsed
0:47:32of images so we don't have any background to
0:47:36there we have learned to do something
0:47:39and then we take that how i look at the conversation goes with the results
0:47:45from human being which was of course a lot of variation
0:47:49because they were absent images
0:47:52this is will be published in icp
0:47:56we sat besides this doctoral dissertation this
0:48:02there are there um many other details yeah in with these applications for the uh
0:48:13forest industry but since i have limited time these i want to the selected so
0:48:20so thank you very much
0:48:40yeah
0:48:51that i
0:48:57i
0:49:03it is difficult to the question and answer is that to my both are important
0:49:09so actually what we finally would like to know is the volume of a of
0:49:15those problems there
0:49:17so then uh we need to detect the problems and but also distribution tells about
0:49:25the process so about been experts can deal when the distribution changes
0:49:32so both are unimportant and of course we make a quite a big assumption here
0:49:37that uh if we think that this is a two dimensional bubble uh you meets
0:49:44tells about the volume button roughly it goes right
0:49:47in liquid
0:50:09uh in this case would actually be are not and we um because these imaging
0:50:15conditions are quite a typical bits totally inside the pipeline and some liquid is run
0:50:20in there and we rely on that fact that at the place where you we
0:50:24are sampling three presents the whole like with the area that might be a little
0:50:29bit risky assumption but uh we uh we had second the possibility of having the
0:50:36uh more images
0:50:57uh
0:51:01yeah
0:51:12um
0:51:14yeah we have a certain number of people first of all form in this uh
0:51:20this uh a human soppy neon and then uh from their we tried to from
0:51:27very low level attributes to build the higher level attributes
0:51:33and the then uh koenig can make them to the physical measurements of is eight
0:51:41samples of since we know each sample and we note no opinion of a human
0:51:46being of from each sample and then uh then be uh for uh yeah
0:51:57start to see the changes um actually it's a it's just send set us up
0:52:02uh some kind of a curve but unfortunately i don't have the curve here we
0:52:08very use it's like a rock or
0:52:10and you check that how close you must goal that yeah this uh but utterance
0:52:15order
0:52:17close to the to the mean opinion score so that is the idea so it's
0:52:23a little bit X experimental you have i understand your question but the that we
0:52:27can say that it's a direct physical uh example exhumation which tells that very good
0:52:35question
0:52:44yeah this one
0:52:52i
0:52:55yeah
0:53:01i
0:53:18um
0:53:20i actually um
0:53:24uh we were of course we must have a normal each of these classes for
0:53:29unsupervised and so what about what about what the band when checking the result of
0:53:34an unsupervised uh
0:53:37clustering but we have to do that without any knowledge of those classes
0:53:47yeah so we have like we have about like uh database and there we have
0:53:51exact knowledge of those categories and we at the end we check how it goes
0:53:58so actually it's a huge the map where all these objects are clustered and then
0:54:05we can start to see how it goes
0:54:09and uh of course results are not so called then in a in a supervised
0:54:14but we were quite surprised about one thing that is that the how a unique
0:54:19the ballot the six people have used like something like uh having twenty of the
0:54:25images and saying something about unsupervised clustering so we used hundreds of maybe six
0:54:32for this
0:54:45oh
0:54:51i
0:55:01actually we are trying to mister those things which affect the quality and actually they
0:55:10affect the usability uh process is there
0:55:16yeah
0:55:21hmmm
0:55:24well as well that's a of it that's a physical actually property of the paper
0:55:29how we cepstral observed scene for example and that's quite the engineering style of testing
0:55:37that there is really putting on the paper and then start into account where other
0:55:41areas it up selves okay and we really doesn't to properly
0:55:47yeah and before it happens that paper then yeah because actually we pretty there is
0:55:56there are not so many technical problems anymore we have a somehow assume that
0:56:02when the paper is okay it's not anymore so pick pretty the quality papers okay
0:56:17ooh
0:56:21i
0:56:24uh excuse me what it what was
0:56:37i mean how well you all these solutions might be what we said difficult to
0:56:42say uh um then uh of course you can calculate the salary of those people
0:56:48work in that club or authorities
0:56:50and sela and the than the price of that uh
0:56:55that solution that machine so that so quite easy to calculate but then when you
0:57:01go to be clear peak approach to cease in paper making or impala been
0:57:07that's a little bit more difficult to calculate but if you see that clear peak
0:57:12lisa in the use of a raw material and water and unity i think the
0:57:17estimates have been even up to forty percent
0:57:23which is a lot of money