i you so i or with the laboratory relay university board a rest and also i'm of where it from a more realistic from tech or shown somebody so what i'm going to present oh these all approach which copes with the particle domain mean namely the animated immediate so the presentation of a line for type i'm going to to the problem state one if is then a pretty state of the art of the to sure proposed proposed approach experimental results and finally conclude the paper so use of that action is um or is part of a more general problem which is temporal segmentation oh you don't because temporal segmentation me it's composing the V don't to its fundamental temporal do needs for be do so a be do so a sequence of images which are four db P of a common or and um so basically to to to get of the final find a movie of the final sequence one has to put to get a all of this short which are are Y what we call gradual transitions which are do not that the image so basically performing the temporal or segmentation means uh on the for basis to be a the be doctrine so we have two classes of we we foundations of for form it's called sharp transitions or cuts which are the direct concatenation two different roles so here you have the time line you have shown one which is connected to show so here i got a car so they are the most frequent for instance a mean of a a bit of for the chip the one cards and the existing approaches i part quite a a highly accurate we got easily and ninety five percent correct detection you can see the the only results of the trick the benchmark mark and compare on the other hand there are the gradual transition which are fourteen time before and the most common we natural movies or or be is in general are face which here i have be give a fate in sequence which is a is the progress a partition of one of each starting with a constant image typically that the other a kind of a idea of trying to are the diesel of each arm much more complex because they are the transformation of one in each start image into two but second image which is done glad so compared to cost they are less frequent at least one word or measure last and the existing methods are not a very high reliable that's say we have a average corner detection between seventy and four for so white white board for means the temporal segmentation i i'm going to an to to to report results so for the on a like this work was it it the way of understanding the structure of the of the be on the other hand we have but the content description for instance the many summarisation a scheme matters are or based on temporal segmentation or oh there are many approaches which can see are the action relate it was high frequency all for change and a to this domain which is the animated movies that use of great you trying to transition has semantic meaning okay so how well i'm going to be then some of the but matters field well that's are with that we a definition of this transition to so supposing would have to sequence is to short S one and S two so that is all transition which is obtained by combining the too of duration they can express it a at intensity level i is the linear combination but between the two seconds with a sequence sorry oh using a do a linear or more point function F one and F some common functions are such as the one i have presented yes so if on a steep because it decreasing for one as you know why the second function F two is typically increasing so basically what we have here we have a a a doll sequence of the for four which is cool we the fading in E C one of the second show so basically we have a fade out cool be the fading C uh oh this kind of time of these of are much more complex to detect compared to the others in one to two face because first of all very hard to to be beat or is a separate uh they they tend to show similar time signature with other channel or or object more based support main evaluation colour X more that that that a and they may have a caught a similar colour is the motion a structure if formation for the whole for the two source of the first one is the can which is a problem so the existing method of equal are divided into several categories of first on it pixel intensity by transform base feature red and there are some other approaches which i don't mixed a fourth one or propose a different solutions so i going to present from each some representative a approach you which are connected to our or oh one of the first approach well you who was using you you in each difference is so i a was to to accumulate the distance between consecutive frames which a should be greater than a of force threshold T one one for the difference for consecutive frames should stay below a second threshold T two which is if you to do you want so basically it the computes the successive difference which are provided by a is all sequence do this work not only for is on but we gradual transmission in general a another approach use the mathematical definition so space that mean and variance of pixel intensity show a linear and quadratic oh behavior so that is find it on on the as a you need we if you are going to compute the variance of what use of sequence once for a different T or want of time we got a quadratic behave or we the F one and F two function so we if you are going to do the mad and replacing the to function we are going to obtain a quadratic behavior according to i so here where a a C R three constants which are in time and keeping in depend oh we can we can uh detect these signature by applying for first or a second or or do but they do but is in order to to do you either a linear decrease or a constant to cost and value of of the of the this fun uh another approach is based on the optical fact i i just my so is a superposition of of fading fade out and in sequence so it detect the amount of fading dean and fading out peaks that which is also the basis for our at forty so generally you you based approaches are very reliable similar to to to the but that's for quite detection other approach are transform base for instance performing forming the detection on the compressed domain this is my work for a real-time performance but that uh the the effect is a quite a visual that we need some kind of visual information not only according for or frequency domain or something similar so usually lead to increase accuracy a least we have to D compressed was that level of detail second and copy what you are feature rate here and going to present a class of one which is based on contour and edging formations so it's use is the same assumption so come to each peak cells from a uh as a starting show are going to disappear why as a can be are from the final four are going to yeah so one classic approach used to compute a edge change ratio for disappearing feature H for edge excels and appearing in edge peaks that for instance that's here we have a the amount of because of quantum piece cells which is that appeared from image at time K divided by the total number of can two points so called my complete do that too they they should should the provide a high value for a for a dissolve other produce that to use feature points like so or see that it's at the top oh the program we feature in for is very sensitive to motion or visual so we do not know the information that the use most in fact all of the existing a dissolve detection method are design actually designed to cope with natural and be because that that was the target so in this paper we address the particular domain mean which is artistic animated movies are not be we stick by a car to ones there are quite a different so and emission mission in the is become a uh that's say an important entertainment in the three from the artistic point of view and also from the entertainment for to there are a lot of it was there are at a or a lot of commercial movie your high i i have used D the of the of the because i state what the law uh cannot up work together and see france for instance the the international house and made at feel more as it's one of the major events in the fields there are a lot of movies competing so a it became a a problem to two to process or from or segmentation to this domain the problem is artistic animated movies are quite different from natural ones in many respects here i'm going to present some of the the most importance so first of one that are many only make animation taking you got paper drawing three D and an object animation blast C modeling so it's the content is very in very different also the motion and not always want you know that you to the animation techniques there are a lot of movies which are made by stop motion take or which are made frame by frame also each movie tend to have a a different colour but i here you have a i i one each or or or two images from a one and with still so they they tend to have a specific colour well that uh that the knees quite fiction or or a highly abstract you have a lot of visual F X job i strange and also there on of physical so we you we can we cannot to unlike the uh the events from the class point to we so basically you can have anything objects appear disappear any kind of visual F X so that is no can oh that is there is no continuous flow so the problem them at the we propose is quite simple but a yet efficient what we do we use only intensity information and for each frame we are going to compute what we call fading excel it the simple racial with the amount of fading out its cells plus the amount of training in excel which is normalized a is back to one of this is a in this size so basically we if we if we are going to a a like this measured you at time shown for use old like uh you isolated peaks the problem is how to make the difference between these all star nation and are changes which are due to motion or visual X so for that we use but between thresholding approach which i shall describe in the form so first of all in order to overcome for all this one you need you we are going to analyse the fading he's than in of very restrained time don't of only three for that is a localisation using that winters for so we have to situation we have a uh that is all which are clearly not which provide a than not a number of fading use so which is quite fight so when whether we have the number of fading be solved a greater than a than a certain threshold and a these value when there is a a lot of i thing we can declare a dissolve in the in there but uh between i and last how to max on the on the on the both sides where T mess is the that's say an average is all a so that the the most simple situation we got oh that is on but there are some other the also which show a lower level of fighting be a and which are cool with all which are put to in other transition like motion or a visual X so we use we use a second trash for which is a quite a lower is lower than the first one we call it the tolerance threshold when are the F B is greater than the second verse what we may have a dissolve transition in fact uh the the frame you made maybe a dissolve middle frame so two to find it is easy is all what we are we are looking for in oh um you know a decreasing in on both sides all this is that so basically having been an mac but what we do here i have think that i uh a that a P function for a a a segment of of of a we we have a to that as a clear is old here and we have the to search for the sort and search for that for that one what but has some other on which are what we some other for change still what we do well we you detected a peak a greater on the second as well we are going to detected time ones where a a if i'm function start increasing the again that on the right and on the left once we got the those times more ones what we we are going to a to assess the and it would be to and the B and those to values which are denoted you left and you die so the transition these value shall be at is on each the to that is are great and then hop the size of the be that the F B I so we are going to be clear that is all okay uh we have tested our uh our approach on of five hundred and S to D all that's several on a midi sequence is for each i have a peak at the a label according to that is the and if you could is that we have a high it difficult content we shall see at the end some examples to as see how how to how bizarre a contents are and average difficulty so to was this perform a we use the class or you don't cold the racial so precision is about false detection while you call is a well-known detection so what are the results so or or one we got a precision of ninety four percent white thirty four is close to eighty percent that you can i sixty good detection and only twenty three for detection but at the sequence level precision and recall racial a range of four at T C to one hundred and the record step one P two one hundred so we have certain second for which we detect all all the mission and there are some for which we we we which you to the very complex and we got to a little or detection issue so we we we have a to compare our of what which is quite simple to the existing approaches so we have to choose three of them the variance of pixel intensities the one i have presented in the introduction okay and the edge change range that they should be um a which is based on two hundred so here we have an example for one movie which is for mister part so we have a trace the the variance of be in T D here we have a D that is on problem to reach he's marked with vertical but lines so we can see that there is no problem shape which is stated by the definition we we can now we can we cannot use it if you are tracing the the exchange ratio we see it it it's very a highly sensitive to visual F X and noise practically unusable usable and if you are a things that the proposed measure we can see whether there are some of duration that is also a quite oh oh that limited and not an example of a we which is the complex as to buy the which show very discontinuous content we got the very which is which show a particular signature what what is not a part shape for green for is not reliable because we don't have a lot of in the movie while that's a classic approach example for our with we get very good all detections so basic we we were unable to compare the precision and recall or for four but approach approach to because we couldn't make them board uh uh i'm going to show you a few examples of a all which were successfully detect a and also to see the difficulty of the of here uh i'm going to show it on a typical dissolve transmission it's quite strange so that's a classic animated movies a but is similar to a fate by is quite a a a quite a diesel if fact here we got a dissolve transition which in which both what will are short are very similar from the point of view of the structure and also the colour and you it it's a tough the the use of trying to which is uh called with a a lot of motion and a very a lot of intensity variation which is also successfully detect so we have proposed an intensity based approach is it's a simple matter is quite a of fast an efficient method to our to or corpus what are the main limitation so forced to one is the choice of of several threshold we had an able to detect of that this all model as you can a channel and we have some probably some of the phase which is reduced or a sometimes as this is you to the the pixels oh thank you for a for hour i think about them any questions have time for one yes to do you mine do but just getting the my a i i just does of forty four okay so for them matter which is this a to information as we use the canny edge detector to know that retrieval okay um another one how how to compute you are a precision and recall um are you considering a and to in which the detection of a describe right or is one single frame but that's that's that would that the good which and so for we are are a menu labelling as the sequence of simple so we are basically detecting by hand uh well that is all are yeah and uh i'm considering reporting detection you find look at yeah yeah i'm support supporting support already the use of it but at least they want people oh we we are not image to detect that is this yeah you you like to so we are so you have to suburb and yeah in in which detection detection it can see does have a someone to what them because can in fact we are not able to detect the one but so we can what to but we can that not problem it was we can uh a statistic we can do that is probably the average is or land for each domain for animated movies was segment three second what for natural reasons maybe twice or that depends on a on a on this domain right thank you very much for example