en-de
en-es
en-fr
en-sl
en
en-zh
0.25
0.5
0.75
1.25
1.5
1.75
2
Learning Similarity Metrics with Invariance Properties
Published on Feb 25, 20079965 Views
Related categories
Chapter list
Learning Similarity Metrics with Invariances00:03
Learning an Invariant Dissimilarity Metric00:21
Dissimilarity Metric for Face Recognition02:36
Siamese Architecture03:54
Dissimilarity Metric vs Traditional Classification04:50
Trainable Metric vs Other Dimensionality Reduction Methods05:10
Trainable Metrics vs handcrafted invariances07:17
Siamese Architecture for Comparing TimeSeries Data07:49
1D Convolutional Net (TDNN)08:41
Examples08:44
Siamese Architecture08:55
Probabilistic Training: Maximum Likelihood09:09
Solution?11:07
Another Loss Function12:23
Loss Function15:31
Examples of Loss Functions16:58
Loss Function: SquareExponential17:26
Face Verification datasets: AT&T, FERET, and AR/Purdue17:37
Face Verification datasets: AT&T, FERET, and AR/Purdue19:04
Face Verification datasets: AT&T, FERET, and AR/Purdue19:25
Face Verification dataset: AR/Purdue19:57
Preprocessing20:04
Centering with a Gaussianblurred face template20:17
Alternated Convolutions and Subsampling20:36
Architecture for the Mapping Function Gw(X)20:41
Internal state for genuine and impostor pairs21:02
Gaussian Face Model in the output space22:25
Dataset for Verification - Verification Results23:05
Classification Examples23:50
lInternal State25:07
DrLim: Dimensionality; Reduction by Learning an Invariant Mapping25:29
“Traditional” Manifold Learning32:10
Learning a FUNCTION from input to output33:17
Learning an INVARIANT FUNCTION from input to output33:24
Learning Invariant Manifolds with EBMs34:32
Step 1: Building a Neighborhood Graph35:29
Step 2: Pick a Parameterized Family of Function36:12
Step 3: Pick a Loss function and Minimize it w.r.t. W36:17
Architecture36:31
Architecture and loss function36:35
Loss function37:08
Mechanical Analogy37:34
MNIST Dataset37:54
MNIST Handwritten Digits. Sanity Check38:01
Architecture of the Gw(X) Function:39:01
Alternated Convolutions and Subsampling39:04
Learning a mapping that is invariant to shifts39:06
Simple Experiment with Shifted MNIST39:27
Shifted MNIST: LLE Result39:42
Shifted MNIST: Injecting Prior Knowledge41:19
Discovering the Viewpoint Manifold42:35
Generic Object Detection and Recognition with Invariance to Pose and Illumination44:14
Data Collection, Sample Generation44:17
NORB Dataset: LLE44:19
Automatic Discovery of the Viewpoint Manifold with Invariant to Illumination45:22
NORB Dataset: Learned Hidden Units46:56