en
0.25
0.5
0.75
1.25
1.5
1.75
2
Learning discriminative space-time actions from weakly labelled videos
Published on 2012-10-093428 Views
Current state-of-the-art action classification methods extract feature representations from the entire video clip in which the action unfolds, however this representation may include irrelevant scen
Related categories
Presentation
Learning Discriminative Space-Time Actions From Weakly Labelled Videos00:00
Outline (1)00:07
Outline (2)00:08
Outline (3)00:18
Outline (4)00:23
Problem definition (1)00:24
Problem definition (2)00:46
Challenging Datasets (1)00:53
Challenging Datasets (2)01:13
Challenging Datasets (3)01:27
Challenging Datasets (4)01:46
State-of-the-art baseline [WUK+09, WKSL11] (1)02:09
State-of-the-art baseline [WUK+09, WKSL11] (2)02:24
State-of-the-art baseline [WUK+09, WKSL11] (3)02:32
State-of-the-art baseline [WUK+09, WKSL11] (4)02:37
State-of-the-art baseline [WUK+09, WKSL11] (5)02:43
State-of-the-art baseline [WUK+09, WKSL11] (6)02:48
Problems with current approach (1)02:54
Problems with current approach (2)03:00
Problems with current approach (3)03:32
Problems with current approach (4)03:53
Problems with current approach (5)03:59
Aim and Contribution (1)04:29
Aim and Contribution (2)04:39
Aim and Contribution (3)05:29
Aim and Contribution (4)05:36
Aim and Contribution (5)05:44
Bag of space-time subvolumes (1)05:50
Max-margin MIL framework (1)06:44
Max-margin MIL framework (2)07:22
Max-margin MIL framework (3)07:31
Max-margin MIL framework (4)07:37
Max-margin MIL framework (5)07:59
Max-margin MIL framework (6)08:04
Max-margin MIL framework (7)08:08
Discriminative action subvolumes (1)08:14
Discriminative action subvolumes (2)08:34
Discriminative action subvolumes (3)08:38
Discriminative action subvolumes (4)08:50
Discriminative action subvolumes (5)08:53
Mapping from instance to bag scores (1)09:05
Mapping from instance to bag scores (2)09:36
Mapping from instance to bag scores (3)09:47
Mapping from instance to bag scores (4)10:01
Mapping from instance to bag scores (5)10:09
Mapping from instance to bag scores (6)10:11
Mapping from instance to bag scores (7)10:13
Experimental set-up (1)10:15
Experimental set-up (2)10:26
Experimental set-up (3)10:33
Results: Video Clip Classification (1)11:19
Results: Video Clip Classification (2)11:28
Results: Video Clip Classification (3)11:34
Results: Video Clip Classification (4)11:44
Results: HOHA2 Video Clip Localisation (1)11:57
Results: HOHA2 Video Clip Localisation (2)12:35
Conclusions & Future Work (1)12:57
Conclusions & Future Work (2)13:05
Conclusions & Future Work (3)13:12
Conclusions & Future Work (4)13:18
Questions?13:38