
en
0.25
0.5
0.75
1.25
1.5
1.75
2
Metric Learning for Large Scale Image Classification: Generalizing to New Classes at Near-Zero Cost
Published on Feb 4, 202523720 Views
We are interested in large-scale image classification and especially in the setting where images corresponding to new or existing classes are continuously added to the training set. Our goal is to dev
Related categories
Presentation
Metric Learning for Large Scale Image Classification: Generalizing to New Classes at Near-Zero Cost00:00
Limitations of 1-vs-Rest SVM30:58:00
k-Nearest Neighbor Classifier - 141:23:48
k-Nearest Neighbor Classifier - 249:59:12
k-Nearest Neighbor Classifier - 350:25:41
Nearest Class Mean Classifier - 152:37:46
Nearest Class Mean Classifier - 259:08:12
Nearest Class Mean Classifier - 362:37:52
Classical: Fisher Discriminant - 165:18:35
Classical: Fisher Discriminant - 269:16:44
Classical: Fisher Discriminant - 374:37:08
Mahalanobis Distance Learning79:56:19
NCM Metric Learning85:52:27
Illustration of Learned Distances97:24:11
Relation to other linear classifiers105:10:05
ILSVRC’10 - Top 5 Accuracy - 1129:21:50
ILSVRC’10 - Top 5 Accuracy - 2137:50:30
Generalization on ImageNet-10K - 1141:22:33
Generalization on ImageNet-10K - 2150:32:33
Transfer Learning - Zero-Shot Prior - 1156:52:38
Transfer Learning - Zero-Shot Prior - 2161:17:44
Transfer Learning - Zero-Shot Prior - 3163:15:39
Transfer Learning - Results ILSVRC’10167:37:44