Capturing Global and Local Dynamics for Human Action Recognition

Size: px
Start display at page:

Download "Capturing Global and Local Dynamics for Human Action Recognition"

Transcription

1 nd Internatonal Conference on Pattern Recognton Capturng Global and Local Dynamcs for Human Acton Recognton Sq Ne Department of Electrcal, Computer and System Engneerng Rensselaer Polytechnc Insttute Troy, New York Qang J Department of Electrcal, Computer and System Engneerng Rensselaer Polytechnc Insttute Troy, New York qj@ecse.rp.edu Abstract Human acton analyss has acheved great success especally wth the recent development of advanced sensors and algorthms that can effectvely track the body jonts. Temporal moton of body jonts carres crucal nformaton about human actons. However, current dynamc models typcally assume statonary local transton and therefore are lmted to local dynamcs. In contrast, we propose a novel human acton recognton algorthm that s able to capture both global and local dynamcs of jont trajectores by combnng a Gaussan-Bnary restrcted Boltzmann machne (GB-RBM) wth a hdden Markov model (HMM). We present a method to use RBM as a generatve model for mult-class classfcaton. Expermental results on benchmark datasets demonstrate the capablty of the proposed method n explotng the dynamc nformaton at dfferent levels. I. INTRODUCTION Human acton s the combnaton of the movements of body jonts over a tme nterval. Understandng a complex acton requres studyng not only the spatal confguratons among the body jonts, but also how they move at dfferent tme scales n the tme doman. Capturng the movements of the body jonts used to be a dffcult task, whch sgnfcantly lmted the performance of prevous vdeo-based human acton recognton, untl the recent emergence of low-cost and relable depth sensors such as Knect and effcent pose trackng systems [18] that can provde well-estmated jont postons n real tme. Jont trajectores present a more explct representaton of the acton dynamcs. However, these temporal characterstcs of human actons have not yet been thoroughly exploted, partally due to the lmtatons of current models. In ths work, we nterpret a human acton as a set of 3D trajectores of domnant body jonts. We comprehensvely nvestgate the underlyng temporal dynamcs of these trajectores for acton recognton. Modelng the temporal patterns of body jonts of a complex human acton s generally addressed by extractng bottom-level spato-temporal features from the mage sequences or desgnng top-level dynamc models such as hdden Markov model (HMM), dynamc Bayesan network (DBN) or condtonal random feld (CRF). Tme-slced dynamc models generally assume n th order Markov property and statonary transton. They, hence, can only capture local statonary transtons but cannot represent global movng pattern. Moreover, these assumptons may not hold for many real-world applcatons. Spato-temporal features are typcally based on local nterest ponts and therefore are also not able to descrbe the movement pattern throughout the whole acton process. Compared to tme-slced dynamc models, restrcted Boltzmann machne (RBM) has been demonstrated to have strong power to capture the jont dstrbuton of the nputs and therefore can be used to model the global patterns when the nput s a tme sequence of jont postons. To the best of our knowledge, RBM has not yet been appled to analyze the global dynamcs of trajectores for acton recognton, although t has been wdely used n many other applcatons such as mage and document analyss. To comprehensvely model the temporal dynamcs of human actons at dfferent levels, we propose a hybrd approach that combnes a Gaussan-Bnary restrcted Boltzmann machne (GB-RBM) to capture the global movement patterns wth an HMM to capture the local dynamcs. As GB-RBM s a varaton of the standard RBM, we use the term RBM n the followng sectons to represent our model. The local and global models capture complementary dynamc nformaton at dfferent tme scales and are combned through a fuson approach for acton classfcaton. A detaled llustraton of the framework s gven n Fgure 1. The remander of the paper s organzed as follows. Secton II presents an overvew of the related work. Secton III ntroduces the learnng process of RBM for acton representaton. Secton IV demonstrates the fuson method for global and local dynamc models. Expermental results are gven n Secton V. The paper s concluded n Secton VI. II. RELATED WORK Human actvty recognton has been wdely nvestgated n the past few decades. Dependng on the acton complexty, human actons are categorzed nto four dfferent levels: gestures, actons, nteractons and group actvtes [1]. In ths work, we focus on acton recognton,.e., a sngle person s actvtes that may be composed of multple gestures organzed temporally, such as wavng, runnng, and jumpng. Research n acton recognton generally follows two paths: sngle-layered approaches and herarchcal approaches. Snglelayered approaches recognze human actons drectly from /14 $ IEEE DOI /ICPR

2 Global Model Tranng Phase Testng Phase Moton Data RBM 1 RBM 2 RBM 3 Acton 1 Acton 2 Acton 3 f 12 f 13 f 23 Parwse Classfcaton Preference Score Model Fuson HMM 1 HMM 2 HMM 3 Local Model Fg. 1. Framework of the proposed method. For each class of acton, one RBM and one HMM are traned to represent the global and local dynamcs respectvely. Each par of RBMs M and M j forms a parwse classfer, whch gves a preference score toward acton class or j. The preference scores of RBM and HMM are combned to make the fnal predcton. sequental mages, whle herarchcal approaches represent actons wth smpler sub-actons. In sngle-layered approaches, sequence of mages may be consdered as 3D volume [16], trajectores [15], or spatotemporal features. The most wdely used spato-temporal features for vsble vdeos are hstogram of gradents (HOG) and hstogram of flows (HOF), whch capture the local appearance or moton nformaton. Features from depth mages and jont trajectores have also been developed recently wth the development of nexpensve and relable depth sensors. For nstance, Wang et al. [20] propose an LOP features whch are the frequency coeffcents of Fourer Transform of local features extracted from the depth mages around human jonts. Gven some specfcally desgned features, template matchng [16], neghborhood-based method [22] and other models are typcally used to make predctons. Herarchcal approaches typcally nclude statstcal approaches and descrpton-based approaches. Statstcal approaches construct statstcal state-based models that are concatenated herarchcally. Condtonal random felds [7] and hdden Markov models [12] are common examples of statstcal models. These dynamc models, ether generatve or dscrmnatve, assume statonary transton and hence are only able to capture local temporal nteractons between several consecutve frames. Descrpton-based approaches dvde human actons nto sub-events. Predcton s made by modelng the temporal and spatal relatonshp of sub-events [6]. Restrcted Boltzmann machne and ts varants are generally used as a tool for feature learnng or data pre-processng yet could also be used for modelng the moton data. For example, Wang et al. [21] uses RBM to get a pror probablty for fnger trace. Larochelle and Bengo [10] uses RBM to generate features for character recognton. Our work s nspred by the dea of Taylor et al. [19], where a Condtonal RBM (CRBM) s proposed to model the temporal transtons between consecutve tme slces and generate pseudo movement sequences. Stll, CRBM models local dynamcs by assumng n th order Markov property. Condtoned on prevous slces, t models the nformaton of the current tme slce. Unlke these works, RBM s used as a generatve model n ths research, whch models the hgh dmensonal sequental data and returns the lkelhood of the nput. Moreover, RBM s combned wth an HMM to jontly capture the global and local dynamcs of human actons. By utlzng an approach to estmate the relatve partton functons of RBMs, we are able to compare between dfferent RBMs, and thus make predctons. III. MODELING TRAJECTORIES In ths work, we propose to capture the global patterns of human jont trajectores usng the restrcted Boltzmann machne (RBM). We choose RBM due to ts capablty to model complex patterns n hgh dmensonal data. One RBM s learned to capture the global movng pattern of one type of acton. In ths secton, we wll frstly gve a bref ntroducton of restrcted Boltzmann machne. We wll then ntroduce how to use RBM to model a sequence of moton data. An approach to estmate the partton functon of RBM s then proposed to perform classfcaton among multple actons. A. Restrcted Boltzmann Machne A restrcted Boltzmann machne (RBM) s a generatve stochastc neural network that can learn a probablty dstrbuton over a set of nputs. As shown n Fgure 2, all the neurons form a bpartte graph: they have nput unts, correspondng to data, hdden unts that are learned, and each connecton n an RBM must connects a vsble unt to a hdden unt. In our work, the hdden unts are bnary and the vsble varables are assumed to follow normal dstrbuton. 1947

3 The energy functon E(v, h) s parameterzed n Equaton 1. Varable a s the bas, σ s the standard devaton of the Gaussan dstrbuton for vsble unt v. If the data s normalzed n each dmenson, then σ =1,a =0. b j s the bas of the hdden unt h j. The jont dstrbuton of the vsble and hdden varables s gven n Equaton 2. Wth contnuous nputs, the partton functon Z can be computed by the ntegral over all vsble nodes and summaton over all hdden unts. E (v, h) = (v a ) 2 2σ 2 v w j h j b j h j, (1) σ j j p (v, h) = 1 exp ( E (v, h)). (2) Z The probablty of an observaton can be calculated by margnalzng over the hdden varables, as shown n Equaton 3. p (v) = 1 exp ( E (v, h)). (3) Z h Hdden unts n the RBM have two states: on and off. Gven an nput vector v, the bnary state h j s set on wth probablty p(h j =1 v) =σ(b j + v w j ), (4) where σ(x) s the logstc sgmod functon 1/(1 + exp( x)). A hdden unt h j s connected to all the nputs, so t s actvated when there exsts some specfc pattern n the vsble layer through Equaton 4. The pattern s captured by the weghts connectng each element n v to h j. Thus the hdden layer h represent mportant patterns of v. Parameters of RBM nclude the weghts of the connectons between the hdden unts and vsble unts as well as ther bases. They are usually learned usng the Contrastve Dvergence (CD) [3] method to get an approxmate Maxmum- Lkelhood soluton. B. Modelng Actons usng RBM Typcally, the nput to the RBM s the lmted to a sngle mage. As we nterpret an acton as a combnaton of the 3D trajectores of human jonts, we propose to use RBM to model the whole sequence of an acton. The basc dea s to feed the jont postons along the temporal trajectory as the nputs to RBM, as shown n Fgure 3, where the t th vsble varable corresponds to the jont postons at tme slce t. Gven a total of N actons, N RBM s {M 1, M 2,, M N } are learned, Hdden Layer wth each model M learnng the temporal dynamcs for acton A. RBM can be effcently learned usng the contrastve dvergence algorthm (CD) [3]. However parameter estmaton of an RBM stll faces one or more of the followng challenges. Due to the non-convexty property of RBM, only local optmal solutons can be acheved. Dfferent ntalzatons could end up wth dfferent estmated parameters. Moreover, parameters are estmated n a generatve manner and therefore t does not necessarly beneft acton classfcaton. We propose a model selecton approach to smultaneously address all the above ssues. Model selecton s performed for every acton n turn. Consder selectng a model for the th acton, we frst generate K canddate RBMs {M k : k = 1,,K} from dfferent ntalzatons. These RBM canddates are then evaluated on the tranng set {V j j }, wth V representng the j th sample of the th acton. The score of each model M s defned n Equaton 5, where E(V j j M) corresponds to the energy of V on model M. The basc dea s that we hope the selected model can maxmally dfferentate the samples of the th acton from other actons n terms of ther lkelhood. Fnally the model that produces the hghest score s selected as the model for the th acton. The procedure s repeated for N tmes untl all the models are selected. Score(M) = j exp( E(V j M)) j exp( E(V j. (5) M)) The dfference between usng energy functon and lkelhood functon s the partton functon. Snce we are computng the lkelhood based on one sngle model, the partton functon s a constant, whch can be omtted n Equaton 5. Wth the RBM learned for each acton, t s nfeasble to compare between models, because calculatng the lkelhood requres calculatng the partton functons, whch s ntractable for RBM wth large number of hdden unts. Nevertheless, there stll exsts a method to estmate the relatve partton functon between dfferent RBM s. For bnary classfcaton, Schmah et al. [17] propose a method to dscrmnatvely estmate the dfference of log-partton functons of two RBMs. t j =logz log Z j. (6) h h 1 2 hm X X X1 2 3 X n Vsble Layer Fg. 2. Graphcal Illustraton of RBM Fg. 3. Modelng Actons wth RBM 1948

4 We extend ths approach to mult-class classfcaton wth a label rankng procedure [9] (Secton IV). C. Local Dynamc Model Local dynamc models capture the local nteractons between consecutve frames. In ths work we mplement hdden Markov model as a local dynamc model. An HMM s defned by the pror of the hdden states, the transton probabltes and the emsson probabltes. The well-known Expectaton-Maxmzaton algorthm (EM) [14] can be employed to estmate the parameters. The hdden states n HMM are generalzaton of the nput sequence. For nstance, f the nputs are actual jont postons, then the hdden states represent some specfc jont postons whch are crucal n the sequence. In ths way a sequence can be transformed nto a sequence of states. To recognze actons, we follow the same procedure as RBM and learn a group of HMM s, each of whch corresponds to one acton. Gven the query sample, ts lkelhood for each HMM s calculated usng the Forward- Backward procedure. HMM s treated as a local dynamc model because t assumes statonary transton and Markov property of the states. We only consder the transton between two consecutve frames. The score of HMM s smply the lkelhood of the observaton, whch s easy to compute usng the Forward-Backward algorthm. IV. FUSION OF GLOBAL AND LOCAL MODELS In ths secton, we ntroduce how we transform unnormalzed lkelhood of RBM nto confdence score, and together wth lkelhood of HMM for acton recognton. A standard procedure to classfy a query sample v s to compute ts lkelhood for all the models, and choose the model wth the greatest lkelhood, as shown n Equaton 7. y =argmaxp(v M ), (7) where y s the predcted result for nstance v. Let p (v) denote the unnormalzed lkelhood n RBM wth log p(v) = logp (v) log Z. A confdence score for a sequence s defned as: 1 F j (v) = 1+exp( α(log p (v M ) log p (v M j ) t j )), (8) where parameter α modfes the dstrbuton of the score, n case all the scores are too close to 0 or 1. The output of such soft bnary classfer can be nterpreted as a confdence value n the classfcaton: the closer the output F j to 1, the stronger the decson of choosng acton A s supported. A valued preference relaton R v s defned for any query nstance v: R v (, j) = { Fj (v) f < j 1 F j (v) f > j. (9) In our approach, we evaluate the score as sum all the confdence value S v () = j R v (, j). (10) The global and local temporal nformaton can be ntegrated at dfferent levels of the learnng process. In ths paper we propose to combne them n the predcton phase. The score of RBM and HMM models are lnearly combned (Equaton 11) wth a tuned weght ω, whch maxmze the recognton accuracy on a valdaton set, and the label wth the hghest score s proposed as the fnal decson. S(v) =S RBM (v)+ωs HMM (v). (11) V. EXPERIMENTS We evaluate our algorthm on three datasets: MSRC- 12 Knect gesture dataset [5], G3D dataset [2], and MSR Acton3D dataset [11]. Models that wll be compared n the experments nclude a global model RBM, a local model HMM, and the combned model. We wll also compare our proposed approach wth other related works. A. MSRC-12 Dataset The Mcrosoft Research Cambrdge-12 Knect gesture dataset conssts of sequences of human movements, represented as body-part locatons, and the assocated gesture to be recognzed by the system. The data set ncludes 594 sequences and 719,359 frames collected from 30 people performng 12 gestures. The moton fles contan tracks of 20 jonts estmated usng the Knect Pose Estmaton ppelne. The body poses are captured at a sample rate of 30Hz wth an accuracy of about two centmeters n jont postons. To deal wth the trackng nose, ansotropc dffuson [13] s employed to smooth the trajectores, correctng nose, yet preservng meanngful changes n moton. Fgure 4 llustrates an example of ansotropc dffuson. As the sze of the vsble layer of an RBM s fxed, lnear nterpolaton s performed to convert all sequences nto the same length (20 frames for each sequence). In ths work, we only use the 3D locaton nformaton of four domnant jonts (.e., two hands and two feet) due to the lmted number of samples. However the proposed approach can be appled to model more jonts f tranng data are adequate. The 3D postons of the body jonts along all three dmensons (x, y and z) are concatenated as the 240-dmenson nput vectors for the RBM model. The sze of hdden layer s set to be 150 accordng to the suggeston n Hnton [8]. The dataset s constructed both to measure the performance of recognton systems and evaluate varous methods of teachng human subjects how to perform dfferent actons. So t s parttoned along dfferent methods of nstructon, such as textonly or text and vdeo. In our work, dfferent nstructons are 1949

5 Fg. 4. Illustraton of the pre-processng of trajectores. The left fgure shows the orgnal trackng result of a jont and the rght fgure shows the processed trajectory after ansotropc dffuson smoothng. It s clear that the hgh frequency nose can be effectvely removed whle the turnng ponts n a trajectory are reserved. LftArms Duck PushRght Goggles WndUp Shoot Bow Throw HadEnough ChangeWeapon Fg. 5. BeatArms Kck LftArms Duck PushRght Goggles WndUp Shoot Bow Throw HadEnough ChangeWeapon BeatArms The confuson matrx of the proposed method on MSRC-12 dataset gnored and only vdeo-based actons are selected to evaluate the performance of the proposed algorthm. The acton markers provded wth the dataset are used to segment the actons from long sequences. 4-fold cross-subject valdaton confguraton s used n our experment. Detaled results are shown n Table I, and the confuson matrx s shown n Fgure 5. The local dynamc model acheves a recognton accuracy of 85.2%, whle the global dynamc model reaches 89.8%. Ths demonstrates the mportance of ncorporatng global dynamcs. Combnng global and local dynamc models, the proposed method can acheve an even better recognton accuracy of 93.1%. In partcular our method outperforms the state-of-the-art method as reported n Ells et al. [4]. Accordng to the confuson matrx, our algorthm performs pretty well on most of the actons, and only fals on a small porton of actons such as Had Enough and Lft Arms. B. G3D Dataset G3D dataset s an acton dataset contanng a range of gamng actons captured by Mcrosoft Knect. The dataset contans 10 subjects performng 20 gamng actons. Synchronzed vdeo, depth and skeleton data are avalable n ths dataset. We only use the skeleton nformaton n the experment. The acton segmentaton s manually labeled. The nput vectors are extracted followng the same procedure n Secton V-A. Half of the samples are used as testng data, 5 samples from each acton as valdaton data, and the other samples as tranng data. To compare wth the baselne method [2], we compute the F1 score for each category of actons. The result s shown n Table II. The proposed method outperforms baselne model for Kck TABLE I. PERFORMANCE COMPARISON OF DIFFERENT METHODS ON MSRC-12 DATASET Method Accuracy Hdden Markov Model 85.2% Ells et al. [4] 88.7% RBM 89.8% Proposed Method 93.1% most of the actons, but also encounters some falures n the actons of Tenns and ThrowBowlngBall. The reason s that when there s occluson of the body parts, the Knect tracker may fal occasonally, and gves the nferred results that wll affect the accuracy, whch s the case of TennsSwngBackhand, Golf and ThrowBowlngBall. Especally n Golf acton, only one sde of the subject can be seen by the camera, so the poston of one leg s nferred usng the trackng procedure, whch brngs n much trouble. Also, the movement range of the acton Walk and Jump s relatvely small, and may be confused wth each other. However, the overall accuracy of our algorthm s acceptable. From Table III, the combned model outperforms the global and local models, as expected. C. MSR Acton3D Dataset MSR Acton3D Dataset s dataset of 20 actons ncludng both depth mages and skeleton trackng results. The dataset reasonably cover the varous movements of arms, legs, torso and ther combnatons. Each acton s performed by 10 subjects, repeated 2 or 3 tmes. There are 567 sequences all together. We use the same four jont postons as our features. Followng the same cross-subject settng as [11], 5 subjects for each acton are selected for testng. For the remanng 5 subjects, 4 are used for tranng and 1 s used for valdaton. Ths dataset poses many challenges for recognton: there exst small between-class varatons (e.g., hgh arm wave and horzontal arm wave), and some actons nvolve complex nteractons among the body parts, thereby leadng to large amount of occlusons (one leg n front of the other or part of body s outsde the camera range) whch sgnfcantly decreases the trackng performance. Table IV llustrates recognton rate of the the local model (HMM), the global model (RBM), the combned approach as well as the results reported n [11], [20]. From the results we can see that the global model outperforms both local models by about 20%. Ths demonstrates the mportance of global dynamcs for dscrmnatng actons, and wth the proposed classfcaton approach, RBM can successfully capture the global dynamcs for acton recognton. Meanwhle, by TABLE II. TABLE III. F1 SCORE OF PROPOSED MODEL AND BASELINE MODEL Acton Bloom et al. [2] Proposed Method Fghtng Golf Tenns Bowlng FPS Drvng Msc Avg COMPARISON OF DIFFERENT METHODS ON G3D DATASET Method Accuracy Hdden Markov Model 77.4% RBM 84.0% Proposed Method 86.4% 1950

6 TABLE IV. COMPARISON OF DIFFERENT METHODS ON ACTION3D DATASET Method Accuracy Hdden Markov Model 55.3% L et al. [11] 74.7& RBM 79.6% Proposed Method 80.2% Wang et al. [20] 88.2% combnng the local model and global model, the proposed method can further mprove the classfcaton accuracy. The performance of our method s below the method reported n [20]. Such result s reasonable because our method only uses a subset of the jonts, and we do not use any features from the depth mages. The feature s much less than other methods that consder both appearance and shape. The proposed method needs more nformaton to classfy among smlar actons such as draw X, draw tck, and draw crcle. The model also cannot handle severely nosy or corrupted data lke bend. However, the proposed algorthm performs qute well on the dstnctve actons, especally complcated actons whch nvolve both hands and feet, lke tenns swng. VI. CONCLUSION In ths paper we propose a novel approach that captures both local and global dynamcal nformaton of the human jont trajectores for acton recognton. The contrbutons of ths paper are as follows. Frst, we ntroduce the Gaussan- Bernoull restrcted Boltzmann machne to model the moton data and capture the global dynamcs of human actons. RBM s used as a generatve model for dynamc modelng. A model selecton method s ntroduced to generate dscrmnatve models. We further propose a novel classfcaton approach to apply RBM for acton recognton. Second, we combne RBM wth hdden Markov model usng a fuson procedure to jontly explot global and local patterns. Fnally, expermental results demonstrate the effectveness of the proposed approach. ACKNOWLEDGMENT The work descrbed n ths paper s supported n part by the grant N from the Offce of Navy Research. [7] L. Han, X. Wu, W. Lang, G. Hou, and Y. Ja. Dscrmnatve human acton recognton n the learned herarchcal manfold space. Image and Vson Computng, 28(5): , [8] G. Hnton. A practcal gude to tranng restrcted boltzmann machnes. Momentum, 9:1, [9] E. Hüllermeer, J. Fürnkranz, W. Cheng, and K. Brnker. Label rankng by learnng parwse preferences. Artfcal Intellgence, 172(16): , [10] H. Larochelle and Y. Bengo. Classfcaton usng dscrmnatve restrcted boltzmann machnes. In Internatonal Conference on Machne Learnng, pages ACM, [11] W. L, Z. Zhang, and Z. Lu. Acton recognton based on a bag of 3d ponts. In Computer Vson and Pattern Recognton Workshops (CVPRW), pages IEEE, [12] F. Lv and R. Nevata. Recognton and segmentaton of 3-d human acton usng hmm and mult-class adaboost. European Conference on Computer Vson (ECCV), pages , [13] P. Perona and J. Malk. Scale-space and edge detecton usng ansotropc dffuson. Pattern Analyss and Machne Intellgence, IEEE Transactons on, 12(7): , [14] L. Rabner. A tutoral on hdden markov models and selected applcatons n speech recognton. Proceedngs of the IEEE, 77(2): , [15] C. Rao and M. Shah. Vew-nvarance n acton recognton. In Computer Vson and Pattern Recognton (CVPR), volume 2, pages II 316. IEEE, [16] M. D. Rodrguez, J. Ahmed, and M. Shah. Acton mach a spatotemporal maxmum average correlaton heght flter for acton recognton. In Computer Vson and Pattern Recognton (CVPR), pages 1 8. IEEE, [17] T. Schmah, G. E. Hnton, S. L. Small, S. Strother, and R. S. Zemel. Generatve versus dscrmnatve tranng of rbms for classfcaton of fmr mages. In Advances n Neural Informaton Processng Systems (NIPS), pages , [18] J. Shotton, A. Ftzgbbon, M. Cook, T. Sharp, M. Fnoccho, R. Moore, A. Kpman, and A. Blake. Real-tme human pose recognton n parts from sngle depth mages. In Computer Vson and Pattern Recognton (CVPR), pages IEEE, [19] G. Taylor, G. Hnton, and S. Rowes. Modelng human moton usng bnary latent varables. Advances n Neural Informaton Processng Systems (NIPS), 19:1345, [20] J. Wang, Z. Lu, Y. Wu, and J. Yuan. Mnng actonlet ensemble for acton recognton wth depth cameras. In Computer Vson and Pattern Recognton (CVPR), pages IEEE, [21] Z. Wang, G. Schalk, and Q. J. Anatomcally constraned decodng of fnger flexon from electrocortcographc sgnals. In Advances n Neural Informaton Processng Systems (NIPS), [22] A. Ylma and M. Shah. Recognzng human actons n vdeos acqured by uncalbrated movng cameras. In Internatonal Conference on Computer Vson (ICCV), volume 1, pages IEEE, REFERENCES [1] J. Aggarwal and M. S. Ryoo. Human actvty analyss: A revew. ACM Computng Surveys (CSUR), 43(3):16, [2] V. Bloom, D. Makrs, and V. Argyrou. G3d: A gamng acton dataset and real tme acton recognton evaluaton framework. In Computer Vson and Pattern Recognton Workshops (CVPRW), pages IEEE, [3] M. Carrera-Perpnan and G. Hnton. On contrastve dvergence learnng. In Artfcal Intellgence and Statstcs, volume 2005, page 17, [4] C. Ells, S. Masood, M. Tappen, J. LaVola, and R. Sukthankar. Explorng the trade-off between accuracy and observatonal latency n acton recognton. Internatonal Journal of Computer Vson, pages 1 17, [5] S. Fothergll, H. M. Ments, P. Kohl, and S. Nowozn. Instructng people for tranng gestural nteractve systems. In J. A. Konstan, E. H. Ch, and K. Höök, edtors, CHI, pages ACM, [6] A. Gupta and L. S. Davs. Objects n acton: An approach for combnng acton understandng and object percepton. In Computer Vson and Pattern Recognton (CVPR), pages 1 8. IEEE,

Outline. Discriminative classifiers for image recognition. Where in the World? A nearest neighbor recognition example 4/14/2011. CS 376 Lecture 22 1

Outline. Discriminative classifiers for image recognition. Where in the World? A nearest neighbor recognition example 4/14/2011. CS 376 Lecture 22 1 4/14/011 Outlne Dscrmnatve classfers for mage recognton Wednesday, Aprl 13 Krsten Grauman UT-Austn Last tme: wndow-based generc obect detecton basc ppelne face detecton wth boostng as case study Today:

More information

Lecture 5: Multilayer Perceptrons

Lecture 5: Multilayer Perceptrons Lecture 5: Multlayer Perceptrons Roger Grosse 1 Introducton So far, we ve only talked about lnear models: lnear regresson and lnear bnary classfers. We noted that there are functons that can t be represented

More information

A Fast Visual Tracking Algorithm Based on Circle Pixels Matching

A Fast Visual Tracking Algorithm Based on Circle Pixels Matching A Fast Vsual Trackng Algorthm Based on Crcle Pxels Matchng Zhqang Hou hou_zhq@sohu.com Chongzhao Han czhan@mal.xjtu.edu.cn Ln Zheng Abstract: A fast vsual trackng algorthm based on crcle pxels matchng

More information

Support Vector Machines

Support Vector Machines /9/207 MIST.6060 Busness Intellgence and Data Mnng What are Support Vector Machnes? Support Vector Machnes Support Vector Machnes (SVMs) are supervsed learnng technques that analyze data and recognze patterns.

More information

Improvement of Spatial Resolution Using BlockMatching Based Motion Estimation and Frame. Integration

Improvement of Spatial Resolution Using BlockMatching Based Motion Estimation and Frame. Integration Improvement of Spatal Resoluton Usng BlockMatchng Based Moton Estmaton and Frame Integraton Danya Suga and Takayuk Hamamoto Graduate School of Engneerng, Tokyo Unversty of Scence, 6-3-1, Nuku, Katsuska-ku,

More information

Classifying Acoustic Transient Signals Using Artificial Intelligence

Classifying Acoustic Transient Signals Using Artificial Intelligence Classfyng Acoustc Transent Sgnals Usng Artfcal Intellgence Steve Sutton, Unversty of North Carolna At Wlmngton (suttons@charter.net) Greg Huff, Unversty of North Carolna At Wlmngton (jgh7476@uncwl.edu)

More information

A Background Subtraction for a Vision-based User Interface *

A Background Subtraction for a Vision-based User Interface * A Background Subtracton for a Vson-based User Interface * Dongpyo Hong and Woontack Woo KJIST U-VR Lab. {dhon wwoo}@kjst.ac.kr Abstract In ths paper, we propose a robust and effcent background subtracton

More information

FEATURE EXTRACTION. Dr. K.Vijayarekha. Associate Dean School of Electrical and Electronics Engineering SASTRA University, Thanjavur

FEATURE EXTRACTION. Dr. K.Vijayarekha. Associate Dean School of Electrical and Electronics Engineering SASTRA University, Thanjavur FEATURE EXTRACTION Dr. K.Vjayarekha Assocate Dean School of Electrcal and Electroncs Engneerng SASTRA Unversty, Thanjavur613 41 Jont Intatve of IITs and IISc Funded by MHRD Page 1 of 8 Table of Contents

More information

Machine Learning 9. week

Machine Learning 9. week Machne Learnng 9. week Mappng Concept Radal Bass Functons (RBF) RBF Networks 1 Mappng It s probably the best scenaro for the classfcaton of two dataset s to separate them lnearly. As you see n the below

More information

Learning the Kernel Parameters in Kernel Minimum Distance Classifier

Learning the Kernel Parameters in Kernel Minimum Distance Classifier Learnng the Kernel Parameters n Kernel Mnmum Dstance Classfer Daoqang Zhang 1,, Songcan Chen and Zh-Hua Zhou 1* 1 Natonal Laboratory for Novel Software Technology Nanjng Unversty, Nanjng 193, Chna Department

More information

Multiple Frame Motion Inference Using Belief Propagation

Multiple Frame Motion Inference Using Belief Propagation Multple Frame Moton Inference Usng Belef Propagaton Jang Gao Janbo Sh The Robotcs Insttute Department of Computer and Informaton Scence Carnege Mellon Unversty Unversty of Pennsylvana Pttsburgh, PA 53

More information

An Image Fusion Approach Based on Segmentation Region

An Image Fusion Approach Based on Segmentation Region Rong Wang, L-Qun Gao, Shu Yang, Yu-Hua Cha, and Yan-Chun Lu An Image Fuson Approach Based On Segmentaton Regon An Image Fuson Approach Based on Segmentaton Regon Rong Wang, L-Qun Gao, Shu Yang 3, Yu-Hua

More information

SLAM Summer School 2006 Practical 2: SLAM using Monocular Vision

SLAM Summer School 2006 Practical 2: SLAM using Monocular Vision SLAM Summer School 2006 Practcal 2: SLAM usng Monocular Vson Javer Cvera, Unversty of Zaragoza Andrew J. Davson, Imperal College London J.M.M Montel, Unversty of Zaragoza. josemar@unzar.es, jcvera@unzar.es,

More information

Tsinghua University at TAC 2009: Summarizing Multi-documents by Information Distance

Tsinghua University at TAC 2009: Summarizing Multi-documents by Information Distance Tsnghua Unversty at TAC 2009: Summarzng Mult-documents by Informaton Dstance Chong Long, Mnle Huang, Xaoyan Zhu State Key Laboratory of Intellgent Technology and Systems, Tsnghua Natonal Laboratory for

More information

A Binarization Algorithm specialized on Document Images and Photos

A Binarization Algorithm specialized on Document Images and Photos A Bnarzaton Algorthm specalzed on Document mages and Photos Ergna Kavalleratou Dept. of nformaton and Communcaton Systems Engneerng Unversty of the Aegean kavalleratou@aegean.gr Abstract n ths paper, a

More information

Fusion Performance Model for Distributed Tracking and Classification

Fusion Performance Model for Distributed Tracking and Classification Fuson Performance Model for Dstrbuted rackng and Classfcaton K.C. Chang and Yng Song Dept. of SEOR, School of I&E George Mason Unversty FAIRFAX, VA kchang@gmu.edu Martn Lggns Verdan Systems Dvson, Inc.

More information

Support Vector Machines

Support Vector Machines Support Vector Machnes Decson surface s a hyperplane (lne n 2D) n feature space (smlar to the Perceptron) Arguably, the most mportant recent dscovery n machne learnng In a nutshell: map the data to a predetermned

More information

Skew Angle Estimation and Correction of Hand Written, Textual and Large areas of Non-Textual Document Images: A Novel Approach

Skew Angle Estimation and Correction of Hand Written, Textual and Large areas of Non-Textual Document Images: A Novel Approach Angle Estmaton and Correcton of Hand Wrtten, Textual and Large areas of Non-Textual Document Images: A Novel Approach D.R.Ramesh Babu Pyush M Kumat Mahesh D Dhannawat PES Insttute of Technology Research

More information

An Entropy-Based Approach to Integrated Information Needs Assessment

An Entropy-Based Approach to Integrated Information Needs Assessment Dstrbuton Statement A: Approved for publc release; dstrbuton s unlmted. An Entropy-Based Approach to ntegrated nformaton Needs Assessment June 8, 2004 Wllam J. Farrell Lockheed Martn Advanced Technology

More information

Detection of an Object by using Principal Component Analysis

Detection of an Object by using Principal Component Analysis Detecton of an Object by usng Prncpal Component Analyss 1. G. Nagaven, 2. Dr. T. Sreenvasulu Reddy 1. M.Tech, Department of EEE, SVUCE, Trupath, Inda. 2. Assoc. Professor, Department of ECE, SVUCE, Trupath,

More information

Edge Detection in Noisy Images Using the Support Vector Machines

Edge Detection in Noisy Images Using the Support Vector Machines Edge Detecton n Nosy Images Usng the Support Vector Machnes Hlaro Gómez-Moreno, Saturnno Maldonado-Bascón, Francsco López-Ferreras Sgnal Theory and Communcatons Department. Unversty of Alcalá Crta. Madrd-Barcelona

More information

Cluster Analysis of Electrical Behavior

Cluster Analysis of Electrical Behavior Journal of Computer and Communcatons, 205, 3, 88-93 Publshed Onlne May 205 n ScRes. http://www.scrp.org/ournal/cc http://dx.do.org/0.4236/cc.205.350 Cluster Analyss of Electrcal Behavor Ln Lu Ln Lu, School

More information

BAYESIAN MULTI-SOURCE DOMAIN ADAPTATION

BAYESIAN MULTI-SOURCE DOMAIN ADAPTATION BAYESIAN MULTI-SOURCE DOMAIN ADAPTATION SHI-LIANG SUN, HONG-LEI SHI Department of Computer Scence and Technology, East Chna Normal Unversty 500 Dongchuan Road, Shangha 200241, P. R. Chna E-MAIL: slsun@cs.ecnu.edu.cn,

More information

Comparing Image Representations for Training a Convolutional Neural Network to Classify Gender

Comparing Image Representations for Training a Convolutional Neural Network to Classify Gender 2013 Frst Internatonal Conference on Artfcal Intellgence, Modellng & Smulaton Comparng Image Representatons for Tranng a Convolutonal Neural Network to Classfy Gender Choon-Boon Ng, Yong-Haur Tay, Bok-Mn

More information

The Research of Support Vector Machine in Agricultural Data Classification

The Research of Support Vector Machine in Agricultural Data Classification The Research of Support Vector Machne n Agrcultural Data Classfcaton Le Sh, Qguo Duan, Xnmng Ma, Me Weng College of Informaton and Management Scence, HeNan Agrcultural Unversty, Zhengzhou 45000 Chna Zhengzhou

More information

Term Weighting Classification System Using the Chi-square Statistic for the Classification Subtask at NTCIR-6 Patent Retrieval Task

Term Weighting Classification System Using the Chi-square Statistic for the Classification Subtask at NTCIR-6 Patent Retrieval Task Proceedngs of NTCIR-6 Workshop Meetng, May 15-18, 2007, Tokyo, Japan Term Weghtng Classfcaton System Usng the Ch-square Statstc for the Classfcaton Subtask at NTCIR-6 Patent Retreval Task Kotaro Hashmoto

More information

Wishing you all a Total Quality New Year!

Wishing you all a Total Quality New Year! Total Qualty Management and Sx Sgma Post Graduate Program 214-15 Sesson 4 Vnay Kumar Kalakband Assstant Professor Operatons & Systems Area 1 Wshng you all a Total Qualty New Year! Hope you acheve Sx sgma

More information

BOOSTING CLASSIFICATION ACCURACY WITH SAMPLES CHOSEN FROM A VALIDATION SET

BOOSTING CLASSIFICATION ACCURACY WITH SAMPLES CHOSEN FROM A VALIDATION SET 1 BOOSTING CLASSIFICATION ACCURACY WITH SAMPLES CHOSEN FROM A VALIDATION SET TZU-CHENG CHUANG School of Electrcal and Computer Engneerng, Purdue Unversty, West Lafayette, Indana 47907 SAUL B. GELFAND School

More information

CS 534: Computer Vision Model Fitting

CS 534: Computer Vision Model Fitting CS 534: Computer Vson Model Fttng Sprng 004 Ahmed Elgammal Dept of Computer Scence CS 534 Model Fttng - 1 Outlnes Model fttng s mportant Least-squares fttng Maxmum lkelhood estmaton MAP estmaton Robust

More information

Problem Definitions and Evaluation Criteria for Computational Expensive Optimization

Problem Definitions and Evaluation Criteria for Computational Expensive Optimization Problem efntons and Evaluaton Crtera for Computatonal Expensve Optmzaton B. Lu 1, Q. Chen and Q. Zhang 3, J. J. Lang 4, P. N. Suganthan, B. Y. Qu 6 1 epartment of Computng, Glyndwr Unversty, UK Faclty

More information

Three supervised learning methods on pen digits character recognition dataset

Three supervised learning methods on pen digits character recognition dataset Three supervsed learnng methods on pen dgts character recognton dataset Chrs Flezach Department of Computer Scence and Engneerng Unversty of Calforna, San Dego San Dego, CA 92093 cflezac@cs.ucsd.edu Satoru

More information

SVM-based Learning for Multiple Model Estimation

SVM-based Learning for Multiple Model Estimation SVM-based Learnng for Multple Model Estmaton Vladmr Cherkassky and Yunqan Ma Department of Electrcal and Computer Engneerng Unversty of Mnnesota Mnneapols, MN 55455 {cherkass,myq}@ece.umn.edu Abstract:

More information

TN348: Openlab Module - Colocalization

TN348: Openlab Module - Colocalization TN348: Openlab Module - Colocalzaton Topc The Colocalzaton module provdes the faclty to vsualze and quantfy colocalzaton between pars of mages. The Colocalzaton wndow contans a prevew of the two mages

More information

Real-time Joint Tracking of a Hand Manipulating an Object from RGB-D Input

Real-time Joint Tracking of a Hand Manipulating an Object from RGB-D Input Real-tme Jont Tracng of a Hand Manpulatng an Object from RGB-D Input Srnath Srdhar 1 Franzsa Mueller 1 Mchael Zollhöfer 1 Dan Casas 1 Antt Oulasvrta 2 Chrstan Theobalt 1 1 Max Planc Insttute for Informatcs

More information

A Fast Content-Based Multimedia Retrieval Technique Using Compressed Data

A Fast Content-Based Multimedia Retrieval Technique Using Compressed Data A Fast Content-Based Multmeda Retreval Technque Usng Compressed Data Borko Furht and Pornvt Saksobhavvat NSF Multmeda Laboratory Florda Atlantc Unversty, Boca Raton, Florda 3343 ABSTRACT In ths paper,

More information

Content Based Image Retrieval Using 2-D Discrete Wavelet with Texture Feature with Different Classifiers

Content Based Image Retrieval Using 2-D Discrete Wavelet with Texture Feature with Different Classifiers IOSR Journal of Electroncs and Communcaton Engneerng (IOSR-JECE) e-issn: 78-834,p- ISSN: 78-8735.Volume 9, Issue, Ver. IV (Mar - Apr. 04), PP 0-07 Content Based Image Retreval Usng -D Dscrete Wavelet wth

More information

Discriminative Dictionary Learning with Pairwise Constraints

Discriminative Dictionary Learning with Pairwise Constraints Dscrmnatve Dctonary Learnng wth Parwse Constrants Humn Guo Zhuoln Jang LARRY S. DAVIS UNIVERSITY OF MARYLAND Nov. 6 th, Outlne Introducton/motvaton Dctonary Learnng Dscrmnatve Dctonary Learnng wth Parwse

More information

Feature Selection for Target Detection in SAR Images

Feature Selection for Target Detection in SAR Images Feature Selecton for Detecton n SAR Images Br Bhanu, Yngqang Ln and Shqn Wang Center for Research n Intellgent Systems Unversty of Calforna, Rversde, CA 95, USA Abstract A genetc algorthm (GA) approach

More information

Classifier Selection Based on Data Complexity Measures *

Classifier Selection Based on Data Complexity Measures * Classfer Selecton Based on Data Complexty Measures * Edth Hernández-Reyes, J.A. Carrasco-Ochoa, and J.Fco. Martínez-Trndad Natonal Insttute for Astrophyscs, Optcs and Electroncs, Lus Enrque Erro No.1 Sta.

More information

Image Representation & Visualization Basic Imaging Algorithms Shape Representation and Analysis. outline

Image Representation & Visualization Basic Imaging Algorithms Shape Representation and Analysis. outline mage Vsualzaton mage Vsualzaton mage Representaton & Vsualzaton Basc magng Algorthms Shape Representaton and Analyss outlne mage Representaton & Vsualzaton Basc magng Algorthms Shape Representaton and

More information

Fast Gesture Recognition with Multiple Stream Discrete HMMs on 3D Skeletons

Fast Gesture Recognition with Multiple Stream Discrete HMMs on 3D Skeletons Fast Gesture Recognton wth Multple Stream Dscrete HMMs on 3D Skeletons Gudo Borgh, Roberto Vezzan and Rta Cucchara DIEF - Unversty of Modena and Reggo Emla Va P. Vvarell 10, 41125 Modena, Italy Emal: {name.surname}@unmore.t

More information

Large-scale Web Video Event Classification by use of Fisher Vectors

Large-scale Web Video Event Classification by use of Fisher Vectors Large-scale Web Vdeo Event Classfcaton by use of Fsher Vectors Chen Sun and Ram Nevata Unversty of Southern Calforna, Insttute for Robotcs and Intellgent Systems Los Angeles, CA 90089, USA {chensun nevata}@usc.org

More information

APPLICATION OF PREDICTION-BASED PARTICLE FILTERS FOR TELEOPERATIONS OVER THE INTERNET

APPLICATION OF PREDICTION-BASED PARTICLE FILTERS FOR TELEOPERATIONS OVER THE INTERNET APPLICATION OF PREDICTION-BASED PARTICLE FILTERS FOR TELEOPERATIONS OVER THE INTERNET Jae-young Lee, Shahram Payandeh, and Ljljana Trajovć School of Engneerng Scence Smon Fraser Unversty 8888 Unversty

More information

Face Detection with Deep Learning

Face Detection with Deep Learning Face Detecton wth Deep Learnng Yu Shen Yus122@ucsd.edu A13227146 Kuan-We Chen kuc010@ucsd.edu A99045121 Yzhou Hao y3hao@ucsd.edu A98017773 Mn Hsuan Wu mhwu@ucsd.edu A92424998 Abstract The project here

More information

Scale Selective Extended Local Binary Pattern For Texture Classification

Scale Selective Extended Local Binary Pattern For Texture Classification Scale Selectve Extended Local Bnary Pattern For Texture Classfcaton Yutng Hu, Zhlng Long, and Ghassan AlRegb Multmeda & Sensors Lab (MSL) Georga Insttute of Technology 03/09/017 Outlne Texture Representaton

More information

Active Contours/Snakes

Active Contours/Snakes Actve Contours/Snakes Erkut Erdem Acknowledgement: The sldes are adapted from the sldes prepared by K. Grauman of Unversty of Texas at Austn Fttng: Edges vs. boundares Edges useful sgnal to ndcate occludng

More information

APPLICATION OF PREDICTION-BASED PARTICLE FILTERS FOR TELEOPERATIONS OVER THE INTERNET

APPLICATION OF PREDICTION-BASED PARTICLE FILTERS FOR TELEOPERATIONS OVER THE INTERNET APPLICATION OF PREDICTION-BASED PARTICLE FILTERS FOR TELEOPERATIONS OVER THE INTERNET Jae-young Lee, Shahram Payandeh, and Ljljana Trajovć School of Engneerng Scence Smon Fraser Unversty 8888 Unversty

More information

Adaptive Silhouette Extraction and Human Tracking in Dynamic. Environments 1

Adaptive Silhouette Extraction and Human Tracking in Dynamic. Environments 1 Adaptve Slhouette Extracton and Human Trackng n Dynamc Envronments 1 X Chen, Zhha He, Derek Anderson, James Keller, and Marjore Skubc Department of Electrcal and Computer Engneerng Unversty of Mssour,

More information

EYE CENTER LOCALIZATION ON A FACIAL IMAGE BASED ON MULTI-BLOCK LOCAL BINARY PATTERNS

EYE CENTER LOCALIZATION ON A FACIAL IMAGE BASED ON MULTI-BLOCK LOCAL BINARY PATTERNS P.G. Demdov Yaroslavl State Unversty Anatoly Ntn, Vladmr Khryashchev, Olga Stepanova, Igor Kostern EYE CENTER LOCALIZATION ON A FACIAL IMAGE BASED ON MULTI-BLOCK LOCAL BINARY PATTERNS Yaroslavl, 2015 Eye

More information

Reducing Frame Rate for Object Tracking

Reducing Frame Rate for Object Tracking Reducng Frame Rate for Object Trackng Pavel Korshunov 1 and We Tsang Oo 2 1 Natonal Unversty of Sngapore, Sngapore 11977, pavelkor@comp.nus.edu.sg 2 Natonal Unversty of Sngapore, Sngapore 11977, oowt@comp.nus.edu.sg

More information

Feature Reduction and Selection

Feature Reduction and Selection Feature Reducton and Selecton Dr. Shuang LIANG School of Software Engneerng TongJ Unversty Fall, 2012 Today s Topcs Introducton Problems of Dmensonalty Feature Reducton Statstc methods Prncpal Components

More information

A New Approach For the Ranking of Fuzzy Sets With Different Heights

A New Approach For the Ranking of Fuzzy Sets With Different Heights New pproach For the ankng of Fuzzy Sets Wth Dfferent Heghts Pushpnder Sngh School of Mathematcs Computer pplcatons Thapar Unversty, Patala-7 00 Inda pushpndersnl@gmalcom STCT ankng of fuzzy sets plays

More information

User Authentication Based On Behavioral Mouse Dynamics Biometrics

User Authentication Based On Behavioral Mouse Dynamics Biometrics User Authentcaton Based On Behavoral Mouse Dynamcs Bometrcs Chee-Hyung Yoon Danel Donghyun Km Department of Computer Scence Department of Computer Scence Stanford Unversty Stanford Unversty Stanford, CA

More information

Real-time Motion Capture System Using One Video Camera Based on Color and Edge Distribution

Real-time Motion Capture System Using One Video Camera Based on Color and Edge Distribution Real-tme Moton Capture System Usng One Vdeo Camera Based on Color and Edge Dstrbuton YOSHIAKI AKAZAWA, YOSHIHIRO OKADA, AND KOICHI NIIJIMA Graduate School of Informaton Scence and Electrcal Engneerng,

More information

Determining the Optimal Bandwidth Based on Multi-criterion Fusion

Determining the Optimal Bandwidth Based on Multi-criterion Fusion Proceedngs of 01 4th Internatonal Conference on Machne Learnng and Computng IPCSIT vol. 5 (01) (01) IACSIT Press, Sngapore Determnng the Optmal Bandwdth Based on Mult-crteron Fuson Ha-L Lang 1+, Xan-Mn

More information

Outline. Type of Machine Learning. Examples of Application. Unsupervised Learning

Outline. Type of Machine Learning. Examples of Application. Unsupervised Learning Outlne Artfcal Intellgence and ts applcatons Lecture 8 Unsupervsed Learnng Professor Danel Yeung danyeung@eee.org Dr. Patrck Chan patrckchan@eee.org South Chna Unversty of Technology, Chna Introducton

More information

Mathematics 256 a course in differential equations for engineering students

Mathematics 256 a course in differential equations for engineering students Mathematcs 56 a course n dfferental equatons for engneerng students Chapter 5. More effcent methods of numercal soluton Euler s method s qute neffcent. Because the error s essentally proportonal to the

More information

Local Quaternary Patterns and Feature Local Quaternary Patterns

Local Quaternary Patterns and Feature Local Quaternary Patterns Local Quaternary Patterns and Feature Local Quaternary Patterns Jayu Gu and Chengjun Lu The Department of Computer Scence, New Jersey Insttute of Technology, Newark, NJ 0102, USA Abstract - Ths paper presents

More information

Simulation: Solving Dynamic Models ABE 5646 Week 11 Chapter 2, Spring 2010

Simulation: Solving Dynamic Models ABE 5646 Week 11 Chapter 2, Spring 2010 Smulaton: Solvng Dynamc Models ABE 5646 Week Chapter 2, Sprng 200 Week Descrpton Readng Materal Mar 5- Mar 9 Evaluatng [Crop] Models Comparng a model wth data - Graphcal, errors - Measures of agreement

More information

Smoothing Spline ANOVA for variable screening

Smoothing Spline ANOVA for variable screening Smoothng Splne ANOVA for varable screenng a useful tool for metamodels tranng and mult-objectve optmzaton L. Rcco, E. Rgon, A. Turco Outlne RSM Introducton Possble couplng Test case MOO MOO wth Game Theory

More information

X- Chart Using ANOM Approach

X- Chart Using ANOM Approach ISSN 1684-8403 Journal of Statstcs Volume 17, 010, pp. 3-3 Abstract X- Chart Usng ANOM Approach Gullapall Chakravarth 1 and Chaluvad Venkateswara Rao Control lmts for ndvdual measurements (X) chart are

More information

Corner-Based Image Alignment using Pyramid Structure with Gradient Vector Similarity

Corner-Based Image Alignment using Pyramid Structure with Gradient Vector Similarity Journal of Sgnal and Informaton Processng, 013, 4, 114-119 do:10.436/jsp.013.43b00 Publshed Onlne August 013 (http://www.scrp.org/journal/jsp) Corner-Based Image Algnment usng Pyramd Structure wth Gradent

More information

Histogram of Template for Pedestrian Detection

Histogram of Template for Pedestrian Detection PAPER IEICE TRANS. FUNDAMENTALS/COMMUN./ELECTRON./INF. & SYST., VOL. E85-A/B/C/D, No. xx JANUARY 20xx Hstogram of Template for Pedestran Detecton Shaopeng Tang, Non Member, Satosh Goto Fellow Summary In

More information

Online Detection and Classification of Moving Objects Using Progressively Improving Detectors

Online Detection and Classification of Moving Objects Using Progressively Improving Detectors Onlne Detecton and Classfcaton of Movng Objects Usng Progressvely Improvng Detectors Omar Javed Saad Al Mubarak Shah Computer Vson Lab School of Computer Scence Unversty of Central Florda Orlando, FL 32816

More information

Adaptive Transfer Learning

Adaptive Transfer Learning Adaptve Transfer Learnng Bn Cao, Snno Jaln Pan, Yu Zhang, Dt-Yan Yeung, Qang Yang Hong Kong Unversty of Scence and Technology Clear Water Bay, Kowloon, Hong Kong {caobn,snnopan,zhangyu,dyyeung,qyang}@cse.ust.hk

More information

Subspace clustering. Clustering. Fundamental to all clustering techniques is the choice of distance measure between data points;

Subspace clustering. Clustering. Fundamental to all clustering techniques is the choice of distance measure between data points; Subspace clusterng Clusterng Fundamental to all clusterng technques s the choce of dstance measure between data ponts; D q ( ) ( ) 2 x x = x x, j k = 1 k jk Squared Eucldean dstance Assumpton: All features

More information

Fuzzy Filtering Algorithms for Image Processing: Performance Evaluation of Various Approaches

Fuzzy Filtering Algorithms for Image Processing: Performance Evaluation of Various Approaches Proceedngs of the Internatonal Conference on Cognton and Recognton Fuzzy Flterng Algorthms for Image Processng: Performance Evaluaton of Varous Approaches Rajoo Pandey and Umesh Ghanekar Department of

More information

Classifier Swarms for Human Detection in Infrared Imagery

Classifier Swarms for Human Detection in Infrared Imagery Classfer Swarms for Human Detecton n Infrared Imagery Yur Owechko, Swarup Medasan, and Narayan Srnvasa HRL Laboratores, LLC 3011 Malbu Canyon Road, Malbu, CA 90265 {owechko, smedasan, nsrnvasa}@hrl.com

More information

Robust Inlier Feature Tracking Method for Multiple Pedestrian Tracking

Robust Inlier Feature Tracking Method for Multiple Pedestrian Tracking 2011 Internatonal Conference on Informaton and Intellgent Computng IPCSIT vol.18 (2011) (2011) IACSIT Press, Sngapore Robust Inler Feature Trackng Method for Multple Pedestran Trackng Young-Chul Lm a*

More information

A Robust Method for Estimating the Fundamental Matrix

A Robust Method for Estimating the Fundamental Matrix Proc. VIIth Dgtal Image Computng: Technques and Applcatons, Sun C., Talbot H., Ourseln S. and Adraansen T. (Eds.), 0- Dec. 003, Sydney A Robust Method for Estmatng the Fundamental Matrx C.L. Feng and Y.S.

More information

Factor Graphs for Region-based Whole-scene Classification

Factor Graphs for Region-based Whole-scene Classification Factor Graphs for Regon-based Whole-scene Classfcaton Matthew R. Boutell Jebo Luo Chrstopher M. Brown CSSE Dept. Res. and Dev. Labs Dept. of Computer Scence Rose-Hulman Inst. of Techn. Eastman Kodak Company

More information

Performance Evaluation of Information Retrieval Systems

Performance Evaluation of Information Retrieval Systems Why System Evaluaton? Performance Evaluaton of Informaton Retreval Systems Many sldes n ths secton are adapted from Prof. Joydeep Ghosh (UT ECE) who n turn adapted them from Prof. Dk Lee (Unv. of Scence

More information

Parallelism for Nested Loops with Non-uniform and Flow Dependences

Parallelism for Nested Loops with Non-uniform and Flow Dependences Parallelsm for Nested Loops wth Non-unform and Flow Dependences Sam-Jn Jeong Dept. of Informaton & Communcaton Engneerng, Cheonan Unversty, 5, Anseo-dong, Cheonan, Chungnam, 330-80, Korea. seong@cheonan.ac.kr

More information

Object-Based Techniques for Image Retrieval

Object-Based Techniques for Image Retrieval 54 Zhang, Gao, & Luo Chapter VII Object-Based Technques for Image Retreval Y. J. Zhang, Tsnghua Unversty, Chna Y. Y. Gao, Tsnghua Unversty, Chna Y. Luo, Tsnghua Unversty, Chna ABSTRACT To overcome the

More information

EECS 730 Introduction to Bioinformatics Sequence Alignment. Luke Huan Electrical Engineering and Computer Science

EECS 730 Introduction to Bioinformatics Sequence Alignment. Luke Huan Electrical Engineering and Computer Science EECS 730 Introducton to Bonformatcs Sequence Algnment Luke Huan Electrcal Engneerng and Computer Scence http://people.eecs.ku.edu/~huan/ HMM Π s a set of states Transton Probabltes a kl Pr( l 1 k Probablty

More information

IEEE TRANSACTIONS ON CIRCUITS AND SYSTEMS FOR VIDEO TECHNOLOGY 1. SSDH: Semi-supervised Deep Hashing for Large Scale Image Retrieval

IEEE TRANSACTIONS ON CIRCUITS AND SYSTEMS FOR VIDEO TECHNOLOGY 1. SSDH: Semi-supervised Deep Hashing for Large Scale Image Retrieval IEEE TRANSACTIONS ON CIRCUITS AND SYSTEMS FOR VIDEO TECHNOLOGY SSDH: Sem-supervsed Deep Hashng for Large Scale Image Retreval Jan Zhang, and Yuxn Peng arxv:607.08477v2 [cs.cv] 8 Jun 207 Abstract Hashng

More information

Action Recognition by Matching Clustered Trajectories of Motion Vectors

Action Recognition by Matching Clustered Trajectories of Motion Vectors Acton Recognton by Matchng Clustered Trajectores of Moton Vectors Mchals Vrgkas 1, Vasleos Karavasls 1, Chrstophoros Nkou 1 and Ioanns Kakadars 2 1 Department of Computer Scence, Unversty of Ioannna, Ioannna,

More information

Action Recognition Using Completed Local Binary Patterns and Multiple-class Boosting Classifier

Action Recognition Using Completed Local Binary Patterns and Multiple-class Boosting Classifier Acton Recognton Usng ompleted Local Bnary Patterns and Multple-class Boostng lassfer Yun Yang, Baochang Zhang, Lnln Yang School of Automaton Scence and Electrcal Engneerng Behang Unversty Beng, hna {yangyun,bczhang,yangln}@buaa.edu.cn

More information

Problem Set 3 Solutions

Problem Set 3 Solutions Introducton to Algorthms October 4, 2002 Massachusetts Insttute of Technology 6046J/18410J Professors Erk Demane and Shaf Goldwasser Handout 14 Problem Set 3 Solutons (Exercses were not to be turned n,

More information

Feature-Based Matrix Factorization

Feature-Based Matrix Factorization Feature-Based Matrx Factorzaton arxv:1109.2271v3 [cs.ai] 29 Dec 2011 Tanq Chen, Zhao Zheng, Quxa Lu, Wenan Zhang, Yong Yu {tqchen,zhengzhao,luquxa,wnzhang,yyu}@apex.stu.edu.cn Apex Data & Knowledge Management

More information

Detection of hand grasping an object from complex background based on machine learning co-occurrence of local image feature

Detection of hand grasping an object from complex background based on machine learning co-occurrence of local image feature Detecton of hand graspng an object from complex background based on machne learnng co-occurrence of local mage feature Shnya Moroka, Yasuhro Hramoto, Nobutaka Shmada, Tadash Matsuo, Yoshak Shra Rtsumekan

More information

Categorizing objects: of appearance

Categorizing objects: of appearance Categorzng objects: global and part-based models of appearance UT Austn Generc categorzaton problem 1 Challenges: robustness Realstc scenes are crowded, cluttered, have overlappng objects. Generc category

More information

Robust visual tracking based on Informative random fern

Robust visual tracking based on Informative random fern 5th Internatonal Conference on Computer Scences and Automaton Engneerng (ICCSAE 205) Robust vsual trackng based on Informatve random fern Hao Dong, a, Ru Wang, b School of Instrumentaton Scence and Opto-electroncs

More information

Compiler Design. Spring Register Allocation. Sample Exercises and Solutions. Prof. Pedro C. Diniz

Compiler Design. Spring Register Allocation. Sample Exercises and Solutions. Prof. Pedro C. Diniz Compler Desgn Sprng 2014 Regster Allocaton Sample Exercses and Solutons Prof. Pedro C. Dnz USC / Informaton Scences Insttute 4676 Admralty Way, Sute 1001 Marna del Rey, Calforna 90292 pedro@s.edu Regster

More information

Dynamic Camera Assignment and Handoff

Dynamic Camera Assignment and Handoff 12 Dynamc Camera Assgnment and Handoff Br Bhanu and Ymng L 12.1 Introducton...338 12.2 Techncal Approach...339 12.2.1 Motvaton and Problem Formulaton...339 12.2.2 Game Theoretc Framework...339 12.2.2.1

More information

Modeling Inter-cluster and Intra-cluster Discrimination Among Triphones

Modeling Inter-cluster and Intra-cluster Discrimination Among Triphones Modelng Inter-cluster and Intra-cluster Dscrmnaton Among Trphones Tom Ko, Bran Mak and Dongpeng Chen Department of Computer Scence and Engneerng The Hong Kong Unversty of Scence and Technology Clear Water

More information

Recognition of Handwritten Numerals Using a Combined Classifier with Hybrid Features

Recognition of Handwritten Numerals Using a Combined Classifier with Hybrid Features Recognton of Handwrtten Numerals Usng a Combned Classfer wth Hybrd Features Kyoung Mn Km 1,4, Joong Jo Park 2, Young G Song 3, In Cheol Km 1, and Chng Y. Suen 1 1 Centre for Pattern Recognton and Machne

More information

High-Boost Mesh Filtering for 3-D Shape Enhancement

High-Boost Mesh Filtering for 3-D Shape Enhancement Hgh-Boost Mesh Flterng for 3-D Shape Enhancement Hrokazu Yagou Λ Alexander Belyaev y Damng We z Λ y z ; ; Shape Modelng Laboratory, Unversty of Azu, Azu-Wakamatsu 965-8580 Japan y Computer Graphcs Group,

More information

Face Tracking Using Motion-Guided Dynamic Template Matching

Face Tracking Using Motion-Guided Dynamic Template Matching ACCV2002: The 5th Asan Conference on Computer Vson, 23--25 January 2002, Melbourne, Australa. Face Trackng Usng Moton-Guded Dynamc Template Matchng Lang Wang, Tenu Tan, Wemng Hu atonal Laboratory of Pattern

More information

Discriminative classifiers for object classification. Last time

Discriminative classifiers for object classification. Last time Dscrmnatve classfers for object classfcaton Thursday, Nov 12 Krsten Grauman UT Austn Last tme Supervsed classfcaton Loss and rsk, kbayes rule Skn color detecton example Sldng ndo detecton Classfers, boostng

More information

Efficient Video Coding with R-D Constrained Quadtree Segmentation

Efficient Video Coding with R-D Constrained Quadtree Segmentation Publshed on Pcture Codng Symposum 1999, March 1999 Effcent Vdeo Codng wth R-D Constraned Quadtree Segmentaton Cha-Wen Ln Computer and Communcaton Research Labs Industral Technology Research Insttute Hsnchu,

More information

2. Related Work Hand-crafted Features Based Trajectory Prediction Deep Neural Networks Based Trajectory Prediction

2. Related Work Hand-crafted Features Based Trajectory Prediction Deep Neural Networks Based Trajectory Prediction Encodng Crowd Interacton wth Deep Neural Network for Pedestran Trajectory Predcton Yanyu Xu ShanghaTech Unversty xuyy2@shanghatech.edu.cn Zhxn Pao ShanghaTech Unversty paozhx@shanghatech.edu.cn Shenghua

More information

Learning a Class-Specific Dictionary for Facial Expression Recognition

Learning a Class-Specific Dictionary for Facial Expression Recognition BULGARIAN ACADEMY OF SCIENCES CYBERNETICS AND INFORMATION TECHNOLOGIES Volume 16, No 4 Sofa 016 Prnt ISSN: 1311-970; Onlne ISSN: 1314-4081 DOI: 10.1515/cat-016-0067 Learnng a Class-Specfc Dctonary for

More information

Pictures at an Exhibition

Pictures at an Exhibition 1 Pctures at an Exhbton Stephane Kwan and Karen Zhu Department of Electrcal Engneerng Stanford Unversty, Stanford, CA 9405 Emal: {skwan1, kyzhu}@stanford.edu Abstract An mage processng algorthm s desgned

More information

Lower Body Pose Estimation in Team Sports Videos Using Label-Grid Classifier Integrated with Tracking-by-Detection

Lower Body Pose Estimation in Team Sports Videos Using Label-Grid Classifier Integrated with Tracking-by-Detection Informaton and Meda Technologes 10(2): 246-258 (2015) reprnted from: IPSJ Transactons on Computer Vson and Applcatons 7: 18-30 (2015) Informaton Processng Socety of Japan Research Paper Lower Body Pose

More information

Efficient Segmentation and Classification of Remote Sensing Image Using Local Self Similarity

Efficient Segmentation and Classification of Remote Sensing Image Using Local Self Similarity ISSN(Onlne): 2320-9801 ISSN (Prnt): 2320-9798 Internatonal Journal of Innovatve Research n Computer and Communcaton Engneerng (An ISO 3297: 2007 Certfed Organzaton) Vol.2, Specal Issue 1, March 2014 Proceedngs

More information

Unsupervised Learning

Unsupervised Learning Pattern Recognton Lecture 8 Outlne Introducton Unsupervsed Learnng Parametrc VS Non-Parametrc Approach Mxture of Denstes Maxmum-Lkelhood Estmates Clusterng Prof. Danel Yeung School of Computer Scence and

More information

Fuzzy Modeling of the Complexity vs. Accuracy Trade-off in a Sequential Two-Stage Multi-Classifier System

Fuzzy Modeling of the Complexity vs. Accuracy Trade-off in a Sequential Two-Stage Multi-Classifier System Fuzzy Modelng of the Complexty vs. Accuracy Trade-off n a Sequental Two-Stage Mult-Classfer System MARK LAST 1 Department of Informaton Systems Engneerng Ben-Guron Unversty of the Negev Beer-Sheva 84105

More information

Gender Classification using Interlaced Derivative Patterns

Gender Classification using Interlaced Derivative Patterns Gender Classfcaton usng Interlaced Dervatve Patterns Author Shobernejad, Ameneh, Gao, Yongsheng Publshed 2 Conference Ttle Proceedngs of the 2th Internatonal Conference on Pattern Recognton (ICPR 2) DOI

More information

A Study on the Application of Spatial-Knowledge-Tags using Human Motion in Intelligent Space

A Study on the Application of Spatial-Knowledge-Tags using Human Motion in Intelligent Space A Study on the Applcaton of Spatal-Knowledge-Tags usng Human Moton n Intellgent Space Tae-Seok Jn*, Kazuyuk Moroka**, Mhoko Ntsuma*, Takesh Sasak*, and Hdek Hashmoto * * Insttute of Industral Scence, the

More information