Home

manual

1. 14 2 2 5 Types of strokes and their meaning 16 Zi Basketball er 2 Aim de thet ae E a A eB A 16 2 3 1 Initializing basketball subset 16 2 3 2 Accessing basketball videos 0 0 16 3 Support 21 Abstract This document briefly describes the CVBASE 2006 dataset which contains videos of sport activity and related data The purpose of the dataset is to provide a sport environment testbed for various methods from the fields of computer vi sion and artificial intelligence Originally the dataset has been prepared in time for the Workshop on Computer Vision Based Analysis in Sport Environments CVBASE 2006 organized in conjunction with ECCV 2006 in Graz Austria The dataset provides three types of data videos player trajectories and expert annotations Included are recordings from three different kinds of sport activity handball squash and basketball While handball and squash data include video trajectories and annotations the basketball data is provided as video only Overall dataset includes approximately 1 hour of video License Basically we do not want unauthorized modifications of the dataset and we do not want dataset to be used in scientific fields for which it was not intended Please read the file license txt which accompanies the dataset for the details We tried to phrase the licence in the way that would give the re searchers maximum freedom of work and publication and
2. Additionally this file contains play phase annotations Matlab function X Y XC YC CVBASEO6SquashGetPos SETNUM PLAYERNUM will extract positions trajectories for the player PLAYERNUM possible values 1 or 2 in the set SETNUM possible values 1 or 2 X Y XC and YC are vectors their length is equal to the number of valid frames frames 1 element per frame X and Y are the positions in the court coordinate system XC and YC are the positions in the camera coordinate system Please note that videos can extend several seconds past the provided player position data those video frames should be discarded Matlab function CVBASEO6SquashShowTrajecties SETNUM PLAYERNUM will show court coordinate system trajectories over a image of squash court Without the second argument it will show trajectories for both players in differ ent colors 2 2 4 Accessing squash annotations Squash dataset provides two types of annotations The first set of annota tions describes phase of the play passive phase vs rally These activities span across certain interval of time This set of annotations is provided as separate column in the tab separated files squash squashipositions txt and squash squash2positions txt Second type of annotations describes player activities and contains the moments when either player strikes a ball These activities have no dura tion They are provided in files squash SquashiPlayerActivity txt and squash Sq
3. All videos with annotations positions and annotation dictionaries are in the tab separated format First row or first and second row in some cases contain names of the columns Dataset contains videos from three types of sports They are only briefly de scribed Users unfamiliar with the rules of the particular sport and the activities that teams and players perform are advised to follow the web links to pages with more detailed explanations 2 1 Handball Handball is a well known indoor sport in Europe and is sometimes referred to as European Handball or Team Handball It should not be confused with the different sport of the same name played mainly in USA Handball is a team sport Each team has 6 players and a goalie a goalkeeper It is played with a ball that is smaller than the ball used in basketball The court is 40 meters long and 20 meters wide The objective of the team is to score points by throwing a ball into the opponents goal Ball is passed between players by throwing by hand There is a 6 meter zone around the goal that players are not allowed to enter with a ball in other words they have to throw before stepping into the area After obtaining a ball either by opponent s goal or mistake the team goes into offensive running across the court After losing a ball team quickly returns to defend its own goal The match lasts one hour two halves 30 minutes each 2 1 Handball 7 More about handb
4. Matlab interface functions reflect that the number of the requested set 1 or 2 has to be passed to most of the functions 2 2 Squash 12 Matlab function CVBASEO6SquashInit PATH String argument PATH should be the path to the root directory of the unpacked dataset Please use the trailing slash or backslash It will load the contents of the following files e squash squash1positions txt and squash squash2positions txt Positions for both of the players for each of the frames from the correspond ing video sequence Columns X1 and Y1 contain the court coordinates of the player 1 and columns X1Cam and Y1Cam contain the positions in the camera coordinates X2 Y2 X2Cam and Y2Cam contain same information for the second player Additionally this file contains the information about the phase of the play for the each frame in the column Phase Value of 1 denotes rally active phase and 2 denotes passive phase To decode phase annotations the following file can be used e squash PhasesDictionary txt Play phases dictionary It could be used to phase numbers back into descriptions e squash SquashiPlayerActivity txt and squash Squash2PlayerActivity txt Player activity annotations Player activities consists of ball strikes including serves Each row contains one strike with the time and frame when the activity was observed The column StrokePlayer contains the number of the player who performed the activity Following
5. encourage wide use of the dataset 1 Data sources This chapter briefly summarizes the sources of video data the process of encod ing transcoding the videos and the sources of annotations 1 1 Handball subset Handball video data was obtained by recording a model handball match not regular league or tournament match Players did not leave the court during the match and did not change their roles during the play Videos annotations and trajectories are from the first 10 minutes of the match 1 1 1 Trajectories player positions Trajectory data was obtained using automatic computer vision based tracking under field expert supervision where possible If automatic tracking did not perform well manual tracking was employed by field expert Positions were entered with variable frequency 1 12 frames interval decision by expert with linear interpolation on the frames between The trajectories from both sources were smoothed by Gaussian kernel 30 30 25 samples Trajectory data is provided for use in higher level processing to provide a shortcut for the researchers who are not able to perform tracking by themselves Trajectory data is supplied both in court coordinates and in the coordinate system of cameras A and B RMS error in player position is approx 0 3 meters directly beneath the camera and approx 0 6 meters at the edge or the middle of the court Do not treat trajectory data as gold standard it is NOT
6. fixed bird s eye view Camera position birds eye view stationary wide angle lens Recorder S VHS videorecorder 1 3 Basketball subset Digitizer Philips DVDR 75 DVD recorder Initial format 25 fps MPEG 2 DVD 720x576 2 fields interlaced Initial bitrate 5 07 MBit s Resized bilinear to 360x288 zero padded to 368x288 Transcoded to DivX 5 2 1 1580 KBit s 1 3 2 HiResBasketballA avi and HiResBasketballB avi Duration 2 minutes each Camera 1 2 PAL CCD camera fixed bird s eye view Camera position birds eye view stationary wide angle lens Recorder S VHS videorecorder Digitizer Philips DVDR 75 DVD recorder Initial format 25 fps MPEG 2 DVD 720x576 2 fields interlaced Initial bitrate 5 07 MBit s Transcoded to DivX 5 2 1 4000 KBit s interlaced 2 Reading and interpreting the data This chapter describes the organization of the dataset including the purpose of the individual files and the ways to load and interpret data Instructions often refer to Matlab functions that are provided for easy access to the data However dataset can be used without Matlab as well if you write your own interface to the provided files All Matlab functions that are used for data access for this dataset start with the prefix CVBASE06 followed by the name of the sport to prevent function name clashes If you use Matlab it may be beneficial to add directory with the dataset related funtions to your Matlab path
7. information is also provided type of the stroke column StrokeType outcome of the stroke StrokeOutcome and the information whether the stroke was forehand or backhand ForehandBackhand column Stroke types are described in more detail at the end of this section Outcome of the stroke is most often the continuation of the play error of the player or point achieved e squash ShotTypeDictionary txt Shot type dictionary If you use Mat lab function for dataset access the annotations will be translated into num bers automatically e squash ShotOutcomeDictionary txt Shot outcome dictionary If you use Matlab function for dataset access the annotations will be translated into numbers automatically e squash ForehandBackhandDictionary txt Shot type forehand or back hand dictionary If you use Matlab function for dataset access the anno tations will be translated into numbers automatically 2 2 Squash 13 2 2 2 Accessing squash videos Videos are stored in the squash directory and are in AVI format compressed with DivX 5 codec see previous chapter for details They can be decompressed using appropriate codecs which are videly available on Internet See the dataset download page http vision fe uni 1j si cvbase06 downloads html for links to programs that are able to convert videos to other formats Matlab function A T CVBASEO6SquashGetFrame SETNUM FRAMENUM will load single frame from the specified squash
8. provided for that purpose and is not sufficiently accurate 1 1 Handball subset 2 1 1 2 Annotations Annotations were supplied by field expert a coach from observations of the video Handbal1C avi They are comprised of two separated sets the first one describes individual player activity annotations passes shots etc and the second one the team activity offense defense etc Annotations and trajectories are supplied just for one of the teams Players of this team were wearing dresses of different colors with the purpose of easy identification Videos A B and C are provided synchronized 1 1 3 HandballA avi and Handball B avi e Duration 10 minutes each e Camera 1 2 PAL CCD camera fixed bird s eye view e Camera position birds eye view stationary wide angle lens e Recorder S VHS videorecorder e Digitizer Pinaclesys DC30 video capture card HW MJPEG e Initial format 25 fps M JPEG 384x576 2 fields interlaced e Initial bitrate 1559 KBit s e Odd fields discarded transcoded to DivX 5 2 1 1580 KBit s 1 1 4 HandballC avi e Duration 10 minutes e Camera Handheld camcorder e Camera position handheld no limitations on motion zoom etc e Recorder Handheld camcorder e Digitizer Pinaclesys DC30 video capture card HW MJPEG e Initial format 25 fps M JPEG 768x576 2 fields interlaced e Initial bitrate 824 KBit s e Odd fields discarded 1 2 reduction in horizontal direction e Transcode
9. the information about the camera used for tracking a particular player on each frame Value of O denotes camera A value of 1 denotes camera B Matlab function X Y XA YA XB YB VALIDCAM CVBASEO06HandballGetPos PLAYERNUM will provide position data the trajectories for the player PLAYERNUM X Y XA YA XB and YB are vectors one element for each frame of video X and Y are the coordinates in court coordinate system XA YA are the coordinates in the image coordinate system of camera A and XB YB are the coordinates in the image coordinate system of camera B VALIDCAM contains information whether camera A or B has been used to track a player at particular frame VALIDCAM value of 0 denotes camera A and 1 denotes camera B Matlab function CVBASEO6HandballShowTrajectories PLAYERNUM will show court coordinate system trajectories over a image of handball court Without argument it will show trajectories for all of the players in different colors 2 1 4 Accessing handball annotations Handball dataset provides two types of annotations The first set of annota tions describes team activity a group activity e g certain type of defense in which whole team participates These activities span across certain in terval of time This set of annotations is provided in the tab separated file handball TeamActivity txt The dictionary which can be used to decode the meaning of annotations and to enumerate them is supplied in the fil
10. when cer tain activity was observed One cannot retrieve player activity by addressing STYPE currentframe Corresponding dictionaries for all annotations are provided in separate files and can be acessed through the following functions Matlab function SHORTDESC LONGDESC CVBASEO6SquashPhasesDict PHASE Matlab function SHORTDESC LONGDESC CVBASEO6SquashShotTypesDict ACT 2 3 Basketball 16 Matlab function SHORTDESC LONGDESC CVBASEO6SquashShotOutcomeDict ACT Matlab function SHORTDESC LONGDESC CVBASEO6SquashShotFBDict ACT 2 2 5 Types of strokes and their meaning The squash subset annotations provide a detailed classification of the type of the stroke for each stroke by either of players Following figures illustrate the meaning of the annotations The path of the ball before hittng the front wall is drawn in solid line The path after hitting the front wall is drawn in dashed line Most of the stroke types have their volley variant which means that the ball did not touch the ground before the stroke 2 3 Basketball Basketball is a well known team sport More about basketball can be found at http en wikipedia org wiki Basketball 2 3 1 Initializing basketball subset When using Matlab interface to the dataset basketball subset must be first ini tialized In initialization phase path to the videos is stored into Matlab memory Since there are no trajectories or annotations pro
11. CVBASE 06 Dataset A Dataset for Development and Testing of Computer Vision Based Methods in Sport Environments User manual and short reference Janez Pers janez pers fe uni lj si Ljubljana October 26 2005 Contents Abstract lv License v 1 Data sources 1 1 1 Ha ndball subsets ce ne een et Ae a ARA 1 1 1 1 Trajectories player positions 1 1 1 2 Annotations 2 1 1 3 HandballA avi and Handball B avi 2 lla Hand baliC ava a a a aa 2 1 2a Sguash subsets araen a ecard Opt tov tHe Gh wel a TR head E 3 1 2 1 Trajectories player positions 3 1 2 2 Annotations A a 3 Lady ESq ashl avi Da AA E 3 Wes BQUASH A AVL A A AA 4 1 3 Basketball subset st nas Dai eh a A AS 4 1 3 1 basketballA avi and basketballB avi 4 1 3 2 HiResBasketballA avi and HiResBasketballB avi 5 2 Reading and interpreting the data 6 Dede Handball bestia tarde y ee Os a rae A Gal a a E Beat Bl dies 6 2 1 1 Initializing handball subset a aoa aaa 7 2 1 2 Accessing handball videos 8 2 1 3 Accessing handball player positions 9 2 1 4 Accessing handball annotations 9 Ded AOS a Bi ee BR Oe A eR et BO A ot 11 2 2 1 Initializing squash subset 6 is 4 4 5 else bane aes 11 2 2 2 Accessing squash videos aooo e a ewes 13 2 2 3 Accessing squash player positions 14 2 2 4 Accessing squash annotations
12. all http en wikipedia org wiki Handball http www usateamhandball org usath sport_basics htm 2 1 1 Initializing handball subset When using Matlab interface to the dataset handball subset must be first initial ized In initialization phase all supplied data except the videos is read into two Matlab variables Subsequent access to trajectories annotations and annotation dictionaries therefore does not require disk access and is quite fast Matlab function CVBASEO6HandballInit PATH String argument PATH should be the path to the root directory of the unpacked dataset Please use the trailing slash or backslash It will load the contents of the following files e handball PlayerList txt List of the players and their numbers tab separated e handball positions txt Positions of the players for each of the frames from the video sequences Players are denoted by their names which correspond to the colors of their dresses for easy visual identification For each player positions are provided in court coordinates and the coordinates of both cameras e handball TeamActivity txt Team activity annotations The format of annotations is as follows first column contains the starting time of team activity in seconds and the second column contains the ending time of ac tivity Whole duration of the video should be covered by team annotations e g when one activity ends the next starts The third column contains ann
13. d to DivX 5 2 1 1580 KBit s 1 2 Squash subset 3 1 2 Squash subset Source of the two squash videos are two different tournaments of different quality levels squash1 avi was recorded on a tournament of recreative players and squash2 avi was recorded at the Slovenian squash championship Both videos were Each video contains one whole set of the squash match In both cases the recorded set was the first set of the match 1 2 1 Trajectories player positions Trajectories provided as part of the squash subset were obtained using auto matic computer vision based tracking under field expert supervision No manual tracking was involved The trajectories were smoothed by Gaussian kernel 30 30 12 samples Trajectory data is provided for use in higher level processing to provide a shortcut for the researchers who are not able to perform tracking by themselves RMS error in player position is approx 0 3 meters Do not treat trajectory data as gold standard it is NOT provided for that purpose and is not sufficiently accurate 1 2 2 Annotations Annotations were provided by sport expert by observing each of the supplied videos They are comprised of two separated sets the first one describes indi vidual player activity annotations strokes types of strokes and the second one describes phase of the play rally or passive phase Stroke annotations also contain position of the ball at the moment of the stroke in came
14. e handball TeamActivityDictionary txt Decoding of annotations is automatic when the provided Matlab functions are used to access the data 2 1 Handball 10 Matlab function TIMESTART TIMEEND FRAMESTART FRAMEEND ACT CVBASEO6HandballGetTeamActivity will return group team activity annotations in vector ACT and the corresponding times seconds and frames when the activity was started and ended are returned in first four vectors Activities are represented by integer numbers which can be translated into descriptions using the following function Matlab function SHORTDESC LONGDESC CVBASEO6HandballTeamDict ACT will return short and long description of annotated team activity supplied as scalar ACT Short description is the annotation which was coined by sport experts to use in annotation files Long description is more informative Matlab function TIME FRAME ACT CVBASEO6HandballGetTeamActivityEx will return group team activity annotations in expanded form Annotations are returned as vector ACT the corresponding times in seconds for each ele ment of ACT is returned in vector TIME and the corresponding frames in vector FRAME NOTE The resulting vectors are of same size than the player position vectors and the number of elements corresponds to the number of video frames in video streams This is essentially the same information than provided by CVBASEO6HandballGetTeamActivity but differently organ
15. ized The second set of annotations describes the activities of individual play ers during the match e g passing a ball throwing at a goal etc These activities are observed in certain instant of time and do not span across in terval of time This set of annotations is provided in the tab separated file handball PlayerActivity txt The dictionary which can be used to de code the meaning of annotations and to enumerate them is supplied in the file handball PlayerActivityDictionary txt Decoding of annotations is auto matic when the provided Matlab functions are used to access the data Matlab function TIME FRAME ACT CVBASEO6HandballGetPlayerActivity PLAYERNUM will return individual player activity annotations in vector ACT and the corre sponding time seconds and frame when the activity was annotated in vectors 2 2 Squash 11 TIME and FRAME Activities are represented by integer numbers which can be translated into descriptions using the following function Matlab function SHORTDESC LONGDESC CVBASEO6HandballPlayerDict ACT will return short and long description of annotated player activity supplied as scalar ACT Short description is the annotation which was coined by sport experts to use in annotation files Long description is more informative 2 2 Squash Squash is a racquet sport played indoors in closed court walls are integral part of the court since the ball bounces off the walls Two p
16. layers are present in a court during the match After the serve the opponent must run to the ball and strike it before bounces off the floor for the second time The ball has to bounce off the front wall inside certain area after the stroke The match is comprised of sets similar as in tennis and the player who wins more sets is declared a winner of the match Each set is comprised of rallies which represent active phase of the play Rally starts with a serve Rally ends when player fails to strike the ball in time and the point or and the serve is awarded to the opponent It can also end for other reasons let is called etc The intervals between the rallies are called passive phases the players generally move slower and prepare for a next rally which will start with a next serve More about squash http en wikipedia org wiki Squash_ sport http www worldsquash co uk rules html 2 2 1 Initializing squash subset When using Matlab interface to the dataset squash subset must be first initial ized In initialization phase all supplied data except the videos is read into two Matlab variables Subsequent access to trajectories annotations and annotation dictionaries therefore does not require disk access and is quite fast Differently from handball subset the two squash videos and the correspond ing annotations are independent of each other They represent different sets in different matches on different tournaments
17. otation mark which can be decoded using the supplied annotation dic tionary e handball TeamActivityDictionary txt Team activity dictionary It should be used to decode annotation marks for team activities Matlab functions perform the translation into numerical codes automatically when the annotation files are loaded e handball PlayerActivity txt Player activity annotations The format of annotations is as follows first column contains the time when activity was 2 1 Handball 8 observed in seconds Second column contains number of the player which performed the activity Player activites do not have any duration e g they are observed at certain time instant The third column contains annotation mark which can be decoded using the supplied annotation dictionary e handball PlayerActivityDictionary txt Player activity dictionary It should be used to decode annotation marks for team activities Matlab functions perform the translation into numerical codes automatically when the annotation files are loaded 2 1 2 Accessing handball videos Videos are stored in the handba11 directory and are in AVI format compressed with DivX 5 codec see previous chapter for details They can be decompressed using appropriate codecs which are videly available on Internet See the dataset download page http vision fe uni 1j si cvbase06 downloads html for links to programs that are able to convert videos to other formats Matlab func
18. ra coordinates This position was entered manually using mouse Due to high speed shutter setting of a camera ball is visible in most of the frames 1 2 3 Squash1 avi e Duration 9 minutes 11 seconds e Camera PAL CCD camera JBL high speed shutter e Camera position birds eye view stationary wide angle lens e Recorder S VHS videorecorder 1 3 Basketball subset 4 Digitizer Pinaclesys DC30 video capture card HW MJPEG Initial format 25 fps M JPEG 384x576 2 fields interlaced Initial bitrate 1504 KBit s Odd fields discarded Transcoded to DivX 5 2 1 1580 KBit s multipass 3 passes 1 2 4 Squash2 avi 1 3 Duration 10 minutes 24 seconds Camera PAL CCD camera JBL high speed shutter Camera position birds eye view stationary wide angle lens Recorder S VHS videorecorder Digitizer Pinaclesys DC30 video capture card HW MJPEG Initial format 25 fps M JPEG 384x576 2 fields interlaced Initial bitrate 1504 KBit s Odd fields discarded Transcoded to DivX 5 2 1 1580 KBit s multipass 3 passes Basketball subset Basketball subset contains videos in two different quality modes Videos depict first few minutes of the play from the slovenian basketball league Higher quality videos shorter overlap with the first partt of the lower quality videos No annotations or trajectory data are provided 1 3 1 basketballA avi and basketball1B avi Duration 5 minutes each Camera 1 2 PAL CCD camera
19. tion A B C T CVBASEO6HandballGetFrames FRAMENUM It will load single frame from each of the handball video streams 3 D arrays A B and C contain truecolor RGB data on return and T contains time seconds at which frames were captured from the frame rate and FRAMENUM Videos A and B were recorded from the birds eye perspective with static camera Video C was recorded from the side and the cameraman was following the action Matlab function CVBASEO6HandballShowFrames FRAMENUM It will load single frame from each of the handball video streams at the frame position FRAMENUM and show all three frames in one figure Player positions will be marked by circles with player names attached to them In many frames players are clearly visible on videos A and B simultaneously Therefore blue circles denote that the other camera has been used for tracking of a particular player and red circle denotes that the current camera has been used 2 1 Handball 9 2 1 3 Accessing handball player positions Player positions are stored in the tab separated file handball positions txt First row contains the names of the players Second row contains the descriptions of individual columns Player positions for each player are given as pairs of coordinates in the coordinate system of the court columns named X and Y in the coordinate system of camera A XA and YA and in the coordinate system of camera B XB and YB Column named Camera gives
20. uash2PlayerActivity txt Matlab function ACT CVBASEO6SquashGetPhasesEx SETNUM 2 2 Squash 15 will return phase rally or passive phase annotations in vector ACT for the set SETNUM NOTE The elements of the resulting vector correspond to the video frames in video streams and to elements of player position vectors Matlab function TIME FRAME STYPE SOUTCOME SFB SX SY CVBASEO6SquashGetPlayerActivity SETNUM PLAYERNUM will return individual player activity annotations strokes Vector STYPE will contain stroke type Decode it using function CVBASEO6SquashShotTypesDict or the corresponding dictionary Squash ShotTypeDictionary txt Vector SOUTCOME will contain stroke outcome Decode it using the function CVBASEO6SquashShotOutcomeDict or the corresponding dictionary Squash ShotOutcomeDictionary txt Vector SFB will contain type of stroke forehand or backhand De code it using function CVBASEO6SquashFBDict or the corresponding dictionary Squash ForehandBackhandDictionary txt Vectors SX and SY will contain coordinates of the ball in the camera co ordinate system at the time the shot was observed Do not mix this with the coordinates of the player at the same time Corresponding time seconds and frame when the activity was annotated is contained in the vectors TIME and FRAME NOTE The resulting vectors of this function are far shorter than when retrieving player trajectories since they contain only the times
21. vided nothing else is loaded Matlab function CVBASEO6BasketballInit PATH String argument PATH should be the path to the root directory of the unpacked dataset Please use the trailing slash or backslash 2 3 2 Accessing basketball videos Dataset provides basketball videos in two formats in the standard quality format e g similar to handball and squash videos and in the high quality for 2 3 Basketball 17 Cross short Drop line amp drop cross less powerful Figure 2 1 Squash stroke types 1 3 2 3 Basketball 18 Boast defensive N Boast offensive Boast reverse Boast skid ball returns to the back Figure 2 2 Squash stroke types 2 3 2 3 Basketball 10 Kill line amp kill cross powerful Serve from the right side Lob line amp lob cross defensive Figure 2 3 Squash stroke types 3 3 2 3 Basketball 20 mat which provides double resolution in both directions and significantly higher bitrate High quality videos are shorter than the low quality counterparts but they originated from the same high quality source Therefore hey entirely overlap with the longer lower quality videos Similar to handball there are two birds eye perspective videos A and B that when combined cover the whole court surface Videos depict single match and are s
22. video s stream Video stream number set 1 or set 2 is specified in SETNUM 3 D array A contains truecolor RGB data T contains time seconds at which frame was captured from framerate and FRAMENUM Matlab function CVBASEO6SquashShowFrame SETNUM FRAMENUM It will load single frame FRAMENUM from the squash video specified by SETNUM Player positions will be marked by circles of different colors If the specified frame contains player activity annotation stroke the position of the ball will be marked with triangle along with other annotation data Try the following example CVBASEO6SquashShowFrame 1 63 Matlab function CVBASEO6SquashReplayActivity SETNUM PLAYERNUM ACTIDX will sequentially load 11 frames from the dataset SETNUM at position defined by PLAYERNUM and ACTIDX and play them on sceen Player positions are marked with circles and position of the ball at the moment of annotation is marked by triangle along with other annotation data ACTIDX is index into the vector of activities e g ACTIDX 1 means first annotated activity for player PLAYERNUM ACTIDX 2 means second etc Try the following example CVBASEO6SquashReplayActivity 1 1 1 2 2 Squash 14 2 2 3 Accessing squash player positions Player positions are stored in the tab separated files squash squashipositions txt and squash squash2positions txt The files contain positions of both of the players both in court and camera coordinate systems
23. ynchronized Matlab function A T A B T CVBASEO6BasketballGetFrames FRAMENUM will load single frame from each of the basketball video streams from the standard quality basketball set 3 D arrays A and B contain truecolor RGB data T contains time seconds at which frames were captured from framerate and FRAMENUM Matlab function A B T CVBASEO6BasketballGetFramesHiRes FRAMENUM will load single frame from each of the basketball video streams from the high quality basketball set 3 D arrays A and B contain truecolor RGB data T contains time seconds at which frames were captured from framerate and FRAMENUM For the difference between the quality settings for both sets of video streams see the first chapter of the manual Matlab function A B T CVBASEO6BasketballShowFrames FRAMENUM will load single frame from the each of the standard quality basketball video streams and display them on screen 3 Support Authors cannot guarantee any kind of support for the provided dataset However if you think that you found an error in the dataset either in the data or in the Matlab code you are welcome to report it to janez pers fe uni 1j si 21

manual

Contents

Download Pdf Manuals

Related Search

Related Contents