Metric Learning for Large Scale Image Classification: Generalizing to New Classes at Near-Zero Cost thumbnail
Pause
Mute
Subtitles
Playback speed
0.25
0.5
0.75
1
1.25
1.5
1.75
2
Full screen

Metric Learning for Large Scale Image Classification: Generalizing to New Classes at Near-Zero Cost

Published on Nov 12, 201223717 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

Chapter list

Metric Learning for Large Scale Image Classification: Generalizing to New Classes at Near-Zero Cost00:00
Motivation00:12
Outline - 100:51
Introduction01:13
Limitations of 1-vs-Rest SVM01:51
Outline - 202:22
k-Nearest Neighbor Classifier - 102:29
k-Nearest Neighbor Classifier - 202:59
k-Nearest Neighbor Classifier - 303:01
Nearest Class Mean Classifier - 103:09
Nearest Class Mean Classifier - 203:32
Nearest Class Mean Classifier - 303:45
Outline - 303:52
Classical: Fisher Discriminant - 103:55
Classical: Fisher Discriminant - 204:09
Classical: Fisher Discriminant - 304:28
Mahalanobis Distance Learning04:47
NCM Metric Learning05:09
Illustration of Learned Distances05:50
Relation to other linear classifiers06:18
Outline - 407:04
Experimental Evaluation07:07
ILSVRC’10 - Top 5 Accuracy - 107:45
ILSVRC’10 - Top 5 Accuracy - 208:16
Generalization on ImageNet-10K - 108:28
Generalization on ImageNet-10K - 209:01
Transfer Learning - Zero-Shot Prior - 109:24
Transfer Learning - Zero-Shot Prior - 209:40
Transfer Learning - Zero-Shot Prior - 309:47
Transfer Learning - Results ILSVRC’1010:03
Outline - 510:54
Conclusion10:56
Thank you11:49