Results 1 - 10 of 42,482

Table 3: Coarse-to-fine multiscale tensor-based motion estimation

in A Coarse To Fine Multiscale Approach For Linear Least Squares Optical Flow Estimation
by F. Lauze, P. Kornprobst, E. Mémin
Cited by 1

Table 7.2 Running time of different parts of the system when coarse-to-fine sampling and lower resolution visibility is used.

in Image-Based Visual Hulls
by Leonard Mcmillan, Wojciech Matusik, Wojciech Matusik

Table 7.2 Running time of different parts of the system when coarse-to-fine sampling and lower resolution visibility is used.

in Title: Assistant Professor
by Wojciech Matusik, Leonard Mcmillan, Wojciech Matusik 2001

Table 6.6 Example 2. Speed-up under a coarse-to-fine mesh sweep.

in MESH-INDEPENDENCE OF SEMISMOOTH NEWTON METHODS FOR LAVRENTIEV-REGULARIZED STATE CONSTRAINED NONLINEAR OPTIMAL CONTROL PROBLEMS
by M. Hintermüller, F. Tröltzsch, I. Yousept

Table 1: Grammar sizes, parsing times and accuracies for hierar- chically split PCFGs with and without hierarchical coarse-to-fine parsing on our development set.

in Learning and Inference for Hierarchically Split PCFGs
by unknown authors
"... In PAGE 4: ... We found our projected grammar estimates to be significantly better suited for pruning than the original grammars, which were learned during training. Experimental Results Table1 shows the tremendous reduction in parsing time (all times are cumulative) and gives an overview over grammar sizes and parsing accuracies. In particular, in our Java im- plementation on a 3GHz processor, it is possible to parse 1600 sentences in less than 900 sec.... ..."

Table 2. Comparison of five classification schemes. The coarse-to-fine scheme gives results similar to the best single-classifier scheme while requiring much less processing.

in large
by Antonin Descampe, Pierre V, Christophe De Vleeschouwer, Benoit Macq

Table 1: Comparisons of the accuracy by various features (%)

in Three Dimensional Continuous DP Algorithm for Multiple
by Sung-phil Heo, Motoyuki Suzuki, Akinori Ito, Shozo Makino
"... In PAGE 2: ...The experimental results for music retrieval are shown in Table1 . To a performance evaluation here, the Coarse-to-Fine and Category 27 method are used (Sonoda, T.... ..."
Cited by 2

Table 1: Comparisons among i) a single SVM dedicated to a small set of hypotheses (in this case a constrained pose domain), ii) the f-network and iii) our designed g-network, for the images in Fig 1. For the single SVM, the position of the face is restricted to a 2 2 window, its scale to the range [10;12] pixels and its orientation to [ 50;+50]; the original image is downscaled 14 times by a factor of 0:83 and for each scale the SVM is applied to the image data around each non-overlapping 2 2 block. In the case of the f and g-networks, we use the coarse-to-fine hierarchy and the search strategy presented here.

in A Hierarchy of Support Vector Machines for Pattern Detection
by Hichem Sahbi, Donald Geman, Pietro Perona 2006
Cited by 3

Table 7: Results of individual action recognition Method Features FER (%) STD

in Modeling Individual and Group Actions in Meetings with Layered HMMs
by Dong Zhang, Daniel Gatica-Perez, Samy Bengio, Iain McCowan, Guillaume Lathoud 2006
"... In PAGE 14: ...resented in [17], we use (0.8,0.2) to weight the audio and visual modalities, respectively. For asynchronous HMM, the allowed asynchrony ranges from a0 a42 a35 a42 a6 . Results are presented in Table7 in terms of FER mean and standard deviation, obtained over 10 runs. From Table 7, we observe that all methods using AV features produced less than 10% FER, which is about 15% absolute improvement over using audio-only features, and about 25% absolute improvement over using visual-only features.... In PAGE 14: ... Results are presented in Table 7 in terms of FER mean and standard deviation, obtained over 10 runs. From Table7 , we observe that all methods using AV features produced less than 10% FER, which is about 15% absolute improvement over using audio-only features, and about 25% absolute improvement over using visual-only features. Asynchronous HMM produced the best result.... ..."
Cited by 7

Table 7: Results of individual action recognition

in Modeling Individual and Group Actions in Meetings: a Two-Layer HMM Framework
by Dong Zhang, Daniel Gatica-perez, Samy Bengio, Iain McCowan, Guillaume Lathoud 2004
"... In PAGE 6: ...hrony ranges from 0.2 seconds to 2.2 seconds. Results are presented in Table7 in terms of FER mean and standard 0-7695-2158-4/04 $20.00 (C) 2004 IEEE ... In PAGE 7: ...50% deviation, obtained over 10 runs. From Table7 , we observe that all methods using AV fea- tures got less than 15% FER, which is about 10% improve- ment over using audio-only features, and about 20% im- provement over using visual-only features. Asynchronous HMM produced the best result.... ..."
Cited by 20
Results 1 - 10 of 42,482