41 Hits in 8.7 sec

Animal pose estimation from video data with a hierarchical von Mises-Fisher-Gaussian model

Libby Zhang, Tim Dunn, Jesse Marshall, Bence Olveczky, Scott W. Linderman
2021 International Conference on Artificial Intelligence and Statistics  
Here we present GIMBAL: a hierarchical von Mises-Fisher-Gaussian model that improves upon deep networks' estimates by leveraging spatiotemporal constraints.  ...  Animal pose estimation from video data is an important step in many biological studies, but current methods struggle in complex environments where occlusions are common and training data is scarce.  ...  Figure 1 : 1 Figure 1: GIMBAL estimates 3D pose from multiple 2D videos with a hierarchical von Mises-Fisher-Gaussian model. a.  ... 
dblp:conf/aistats/ZhangDMOL21 fatcat:u454i5hbpjgipjv3g4ipa7dbk4


Manolis Savva, Angel X. Chang, Pat Hanrahan, Matthew Fisher, Matthias Nießner
2016 ACM Transactions on Graphics  
Using this pose and 3D scene data we learn a set of prototypical interaction graphs (PiGraphs) containing priors on the human pose and object geometry during interaction (middle).  ...  Write on a whiteboard Training Data Learned PiGraphs Generated Interaction Snapshots Sit on a chair and watch TV Sit on a couch and use a laptop whiteboard chair TV TV stand Figure 1: We use commodity  ...  Acknowledgements We gratefully acknowledge the support of NVIDIA Corporation with the donation of Titan GPUs used for this research. This work is partially supported by NSF Award IIS-1514268.  ... 
doi:10.1145/2897824.2925867 fatcat:p7oytlqrzbcb3iugcvdmo3dwmy

VAE-SNE: a deep generative model for simultaneous dimensionality reduction and clustering [article]

Jacob M. Graving, Iain D. Couzin
2020 bioRxiv   pre-print
Finally, we also introduce variants of VAE-SNE for embedding data in polar (spherical) coordinates and for embedding image data from raw pixels.  ...  Our model combines elements from deep learning, probabilistic inference, and manifold learning to produce interpretable compressed representations while also readily scaling to tens-of-millions of observations  ...  S Spherical embeddings with von Mises-Fisher VAE-SNE. a-b, Spherical embeddings using VAE-SNE with a von Mises-Fisher similarity kernel (Appendix C. ) of the posture dynamics dataset (a; Video S8) from  ... 
doi:10.1101/2020.07.17.207993 fatcat:x4g6qwa62bcdznfxfjoeorv4zy

Machine learning in acoustics: theory and applications [article]

Michael J. Bianco, Peter Gerstoft, James Traer, Emma Ozanich, Marie A. Roch, Sharon Gannot, Charles-Alban Deledalle
2019 arXiv   pre-print
With large volumes of training data, ML can discover models describing complex acoustic phenomena such as human speech and reverberation.  ...  Acoustic data provide scientific and engineering insights in fields ranging from biology and communications to ocean and Earth science.  ...  model. 124 n another variant of the classification paradigm, the GMM is substituted by a mixture of von Mises, 125 which is a suitable distribution for the periodic phase of the microphone signals.  ... 
arXiv:1905.04418v3 fatcat:xuhykqhrk5bqbg5x7gaajcksay

Image splicing detection with local illumination estimation

Yu Fan, Philippe Carre, Christine Fernandez-Maloigne
2015 2015 IEEE International Conference on Image Processing (ICIP)  
PATHOLOGY IMAGESVIA VON MISES MIXTURE MODEL Xingyu LI, University of Toronto Konstantinos N .  ...  super-resolution models where examples are learned from one or more scales of the input data .  ...  The sensory gap will then be the focus, and its causes (e .g ., resolution, perspective, scale, field of view) will be highlighted with a visual demonstration .  ... 
doi:10.1109/icip.2015.7351341 dblp:conf/icip/FanCF15 fatcat:7ja5gjnp5rafvedc2nman7xcru

State of the art in photon density estimation

Toshiya Hachisuka, Andrew Selle, Ben Spencer, Wojciech Jarosz, Guillaume Bouchard, Per Christensen, Jeppe Revall Frisvad, Wenzel Jakob, Henrik Wann Jensen, Michael Kaschalk, Claude Knaus
2012 ACM SIGGRAPH 2012 Posters on - SIGGRAPH '12  
demo: • Run on an NVidia GeForce 9600M GT graphics card Limitations and future work • Anisotropic media currently not supported potential extension using von Mises-Fisher distributions •  ...  One option is to per- form the fitting procedure entirely in a 5D spatio-directional domain where each component is the tensor product of a spa- tial Gaussian and a von Mises-Fischer distribution.  ... 
doi:10.1145/2343483.2343489 fatcat:lmmwllfeyjd5tibrm3yzkntcly

2013 Index IEEE Transactions on Biomedical Engineering Vol. 60

2013 IEEE Transactions on Biomedical Engineering  
-that appeared in this periodical during 2013, and items from previous years that were commented upon or corrected in 2013.  ...  Shibanoki, Quaternionic Attitude Estimation for Robotic and Human Motion Tracking Using Sequential Monte Carlo Methods With von Mises-Fisher and Nonuniform Densities Simulations.  ...  R., Quaternionic Attitude Estimation for Robotic and Human Motion Tracking Using Sequential Monte Carlo Methods With von Mises-Fisher and Nonuniform Densities Simulations; TBME Nov. 2013 3046-3059 Tokuda  ... 
doi:10.1109/tbme.2013.2295159 fatcat:kjadiysea5h55hjr7i3o6vstlm

On Fitting Finite Dirichlet Mixture Using ECM and MML [chapter]

Nizar Bouguila, Djemel Ziou
2005 Lecture Notes in Computer Science  
Fluid queue models play a vital role in the performance analysis of computer and telecommunication networks.  ...  We study a fluid system with single on/off source in which the active (on) and silent periods (off) follow General and Exponential distributions respectively.  ...  We study non-Gaussian weak limits for canonical U-statistics (and for canonical Von Mises statistics as well) based on samples from stationary sequences satisfying α or ϕ-mixing conditions. Theorem.  ... 
doi:10.1007/11551188_19 fatcat:gtzi5u6emvfyxea3zbklobnvom

A massive 7T fMRI dataset to bridge cognitive and computational neuroscience [article]

Emily J. Allen, Ghislain St-Yves, Yihan Wu, Jesse L. Breedlove, Logan T. Dowdle, Brad Caron, Franco Pestilli, Ian Charest, J. Benjamin Hutchinson, Thomas Naselaris, Kendrick Kay
2021 bioRxiv   pre-print
To optimize data quality, we develop and apply novel estimation and denoising techniques.  ...  Further exemplifying the inferential power of the dataset, we use NSD to build and train deep neural network models that predict brain activity more accurately than state-of-the-art models from computer  ...  a mixing ratio of 60% and 40% for the von Mises and uniform 30 distributions, respectively.  ... 
doi:10.1101/2021.02.22.432340 fatcat:2vrfrcntmbgphky563lxig5hn4

ASE: Large-Scale Reusable Adversarial Skill Embeddings for Physically Simulated Characters [article]

Xue Bin Peng, Yunrong Guo, Lina Halper, Sergey Levine, Sanja Fidler
2022 arXiv   pre-print
This is in stark contrast to what is common practice in physics-based character animation, where control policies are most typically trained from scratch for each task.  ...  We show that a single pre-trained model can be effectively applied to perform a diverse set of new tasks.  ...  ACKNOWLEDGMENTS We would like to thank Reallusion 1 for providing reference motion data for this project, the anonymous reviewers for their feedback, Charles Zhou, Dane Johnston, David Dixon, Simon Yuen  ... 
arXiv:2205.01906v1 fatcat:th7kbskg4zc2bjjh3jvkrldfby

Deep Sequence Learning for Video Anticipation: From Discrete and Deterministic to Continuous and Stochastic [article]

Sadegh Aliakbarian
2020 arXiv   pre-print
In this research, we start from anticipating a coarse representation of a deterministic future and then move towards predicting continuous and fine-grained future representations of a stochastic process  ...  Based on the nature of the task, video anticipation can be considered from two viewpoints: the level of details and the level of determinism in the predicted future.  ...  KL divergence equal to log K, with K the size of the codebook; several recent works use the von Mises-Fisher distribution to obtain a fixed KL divergence, thus mitigating the posterior collapse problem  ... 
arXiv:2010.04368v1 fatcat:vip2fzkp3becxginqlioskbrfy

Irregular distribution of grid cell firing fields in rats exploring a 3D volumetric space

Roddy M. Grieves, Selim Jedidi-Ayoub, Karyna Mishchanchuk, Anyi Liu, Sophie Renaudineau, Éléonore Duvelle, Kate J. Jeffery
2021 Nature Neuroscience  
On a horizontal surface, grid cells usually produce multiple, spatially focal, approximately circular firing fields that are evenly sized and spaced to form a regular, close-packed, hexagonal array.  ...  In three dimensions, theoretically the equivalent firing pattern would be a regular, hexagonal close packing of evenly sized spherical fields.  ...  This work was supported by grants from Wellcome (grant nos. 103896AIA to K.J., 208647/Z/17/Z to S.J-A. and 213295/Z/18/Z to A.L.).  ... 
doi:10.1038/s41593-021-00907-4 pmid:34381241 pmcid:PMC8553607 fatcat:3tnkg5chifh3lpignul7vwzkza

Multiple object tracking with context awareness [article]

Laura Leal-Taixé
2016 arXiv   pre-print
Multiple people tracking is a key problem for many applications such as surveillance, animation or car navigation, and a key input for tasks such as activity recognition.  ...  coming from different views.  ...  Outdoor human motion capture using inverse kinematics and von Mises-Fisher sampling. IEEE International Conference on Computer Vision (ICCV), November 2011. [ 17 ] 17 G. Pons-Moll, L.  ... 
arXiv:1411.7935v2 fatcat:hv4ge7lgubhnvcbuwibg3kmhia

A combined local and global motion estimation and compensation method for cardiac CT

Qiulin Tang, Beshan Chiang, Akinola Akinyemi, Alexander Zamyatin, Bibo Shi, Satoru Nakanishi, Bruce R. Whiting, Christoph Hoeschen
2014 Medical Imaging 2014: Physics of Medical Imaging  
First, random undersampling masks were generated based on Gaussian, Cauchy, wrapped Cauchy and von Mises probability distribution functions by the inverse transform method.  ...  We used Gaussian Mixture Model with Expectation Maximization (GMM-EM), (EM is used to estimate the model parameters) to automatically separate FDT data into clusters of normal and abnormal eyes.  ...  FTIR chemical imaging allows the spatial distribution of chemistry to be rapidly imaged at a high (diffraction limited) spatial resolution where a pixel represents an area of 5.5 ? 5.5 µm2 of tissue.  ... 
doi:10.1117/12.2043492 fatcat:fyzpc5m6jbh7fjohqpdmtzkhte

Where in the World is this Image? Transformer-based Geo-localization in the Wild [article]

Shraman Pramanick, Ewa M. Nowara, Joshua Gleason, Carlos D. Castillo, Rama Chellappa
2022 arXiv   pre-print
Predicting the geographic location (geo-localization) from a single ground-level RGB image taken anywhere in the world is a very challenging problem.  ...  a single image possibly having only a few geo-locating cues.  ...  •MvMF[23]introducestheMixture of von-Mises Fisher (MvMF) loss function for the classification layer that exploits the earth's spherical geometry and refines the geographical cell shapes in the partitioning  ... 
arXiv:2204.13861v2 fatcat:uphqixmbxjg4baxwja6jdoqbee
« Previous Showing results 1 — 15 out of 41 results