weakly supervised learning from images and videolaptev/share/slides/hackathon17... · 2017. 2....
TRANSCRIPT
![Page 1: Weakly supervised learning from images and videolaptev/share/slides/hackathon17... · 2017. 2. 15. · Ivan Laptev ivan.laptev@inria.fr WILLOW, INRIA/ENS/CNRS, Paris Weakly supervised](https://reader033.vdocuments.us/reader033/viewer/2022060600/6053d5ffcbad1e388c6b1557/html5/thumbnails/1.jpg)
Ivan [email protected]
WILLOW, INRIA/ENS/CNRS, Paris
Weakly supervised learning
from images and video
See.4C Spatio-temporal Series Hackathon
February 14, 2017
Joint work with: Maxime Oquab – Piotr Bojanowski – Rémi Lajugie –
Jean-Baptiste Alayrac – Leon Bottou – Francis Bach –
Simon Lacoste-Julien – Jean Ponce – Cordelia Schmid – Josef Sivic
![Page 2: Weakly supervised learning from images and videolaptev/share/slides/hackathon17... · 2017. 2. 15. · Ivan Laptev ivan.laptev@inria.fr WILLOW, INRIA/ENS/CNRS, Paris Weakly supervised](https://reader033.vdocuments.us/reader033/viewer/2022060600/6053d5ffcbad1e388c6b1557/html5/thumbnails/2.jpg)
What is Computer Vision?
![Page 3: Weakly supervised learning from images and videolaptev/share/slides/hackathon17... · 2017. 2. 15. · Ivan Laptev ivan.laptev@inria.fr WILLOW, INRIA/ENS/CNRS, Paris Weakly supervised](https://reader033.vdocuments.us/reader033/viewer/2022060600/6053d5ffcbad1e388c6b1557/html5/thumbnails/3.jpg)
Computer vision works
![Page 4: Weakly supervised learning from images and videolaptev/share/slides/hackathon17... · 2017. 2. 15. · Ivan Laptev ivan.laptev@inria.fr WILLOW, INRIA/ENS/CNRS, Paris Weakly supervised](https://reader033.vdocuments.us/reader033/viewer/2022060600/6053d5ffcbad1e388c6b1557/html5/thumbnails/4.jpg)
Recent Progress:
Convolutional Neural Networks
2012:
VGG: 6.8%
GoogLeNet: 6.6%
BAIDU 5.3%
Human 5.1%
ResNet 3.6%
ILSVRC’12: 1.2M images, 1K classes
Top 5 error:
Object classification
2014-2015:
Face Recognition
LFW
Sam
eD
iffe
rent
DeepFace 97.3%
VGG 99.1%
Human 99.2%
VisionLabs 99.3%
FaceNet 99.6%
BAIDU 99.7%
Accuracy:
2014-2016:
--2013:LBP 87.3%
FVF 93.0%
![Page 5: Weakly supervised learning from images and videolaptev/share/slides/hackathon17... · 2017. 2. 15. · Ivan Laptev ivan.laptev@inria.fr WILLOW, INRIA/ENS/CNRS, Paris Weakly supervised](https://reader033.vdocuments.us/reader033/viewer/2022060600/6053d5ffcbad1e388c6b1557/html5/thumbnails/5.jpg)
How does it work?
AlexNet [Krizhevsky et al. 2012]
~60M parameters
Image
annotation
![Page 6: Weakly supervised learning from images and videolaptev/share/slides/hackathon17... · 2017. 2. 15. · Ivan Laptev ivan.laptev@inria.fr WILLOW, INRIA/ENS/CNRS, Paris Weakly supervised](https://reader033.vdocuments.us/reader033/viewer/2022060600/6053d5ffcbad1e388c6b1557/html5/thumbnails/6.jpg)
Problems with annotation
Expensive
Ambiguous
Table? Dining table? Desk? …
![Page 7: Weakly supervised learning from images and videolaptev/share/slides/hackathon17... · 2017. 2. 15. · Ivan Laptev ivan.laptev@inria.fr WILLOW, INRIA/ENS/CNRS, Paris Weakly supervised](https://reader033.vdocuments.us/reader033/viewer/2022060600/6053d5ffcbad1e388c6b1557/html5/thumbnails/7.jpg)
Problems with annotationWhat action class?
![Page 8: Weakly supervised learning from images and videolaptev/share/slides/hackathon17... · 2017. 2. 15. · Ivan Laptev ivan.laptev@inria.fr WILLOW, INRIA/ENS/CNRS, Paris Weakly supervised](https://reader033.vdocuments.us/reader033/viewer/2022060600/6053d5ffcbad1e388c6b1557/html5/thumbnails/8.jpg)
Problems with annotationWhat action class?
![Page 9: Weakly supervised learning from images and videolaptev/share/slides/hackathon17... · 2017. 2. 15. · Ivan Laptev ivan.laptev@inria.fr WILLOW, INRIA/ENS/CNRS, Paris Weakly supervised](https://reader033.vdocuments.us/reader033/viewer/2022060600/6053d5ffcbad1e388c6b1557/html5/thumbnails/9.jpg)
How to avoid manual supervision?
Weakly-supervised learning from
images and video
![Page 10: Weakly supervised learning from images and videolaptev/share/slides/hackathon17... · 2017. 2. 15. · Ivan Laptev ivan.laptev@inria.fr WILLOW, INRIA/ENS/CNRS, Paris Weakly supervised](https://reader033.vdocuments.us/reader033/viewer/2022060600/6053d5ffcbad1e388c6b1557/html5/thumbnails/10.jpg)
pre-train CNN
on ImageNet
[Girshick’15], [Girshick et al.’14], [Oquab et al.’14], [Sermanet et al.’13 ], [Donahue et al. ’13],
[Zeiler & Fergus ’13] ...
Train CNNs for object detection
FCa
C1-C2-C3-C4-C5 FC6 FC7
FCa
chairchairperson
backgr.
●●● table
Convolutional layersFully
Connected
layers
![Page 11: Weakly supervised learning from images and videolaptev/share/slides/hackathon17... · 2017. 2. 15. · Ivan Laptev ivan.laptev@inria.fr WILLOW, INRIA/ENS/CNRS, Paris Weakly supervised](https://reader033.vdocuments.us/reader033/viewer/2022060600/6053d5ffcbad1e388c6b1557/html5/thumbnails/11.jpg)
Results
Oquab, Bottou, Laptev and Sivic
CVPR 2014
Pascal VOC
![Page 12: Weakly supervised learning from images and videolaptev/share/slides/hackathon17... · 2017. 2. 15. · Ivan Laptev ivan.laptev@inria.fr WILLOW, INRIA/ENS/CNRS, Paris Weakly supervised](https://reader033.vdocuments.us/reader033/viewer/2022060600/6053d5ffcbad1e388c6b1557/html5/thumbnails/12.jpg)
[Oquab, Bottou, Laptev and Sivic, CVPR 2014]
Results
![Page 13: Weakly supervised learning from images and videolaptev/share/slides/hackathon17... · 2017. 2. 15. · Ivan Laptev ivan.laptev@inria.fr WILLOW, INRIA/ENS/CNRS, Paris Weakly supervised](https://reader033.vdocuments.us/reader033/viewer/2022060600/6053d5ffcbad1e388c6b1557/html5/thumbnails/13.jpg)
FCa
C1-C2-C3-C4-C5 FC6 FC7
FCa
chairchairperson
backgr.
●●● table
Convolutional layersFully
Connected
layers
Problem: Annotation of bounding boxes is (a): expensive (b): subjective
How to use CNNs for cluttered scenes?
![Page 14: Weakly supervised learning from images and videolaptev/share/slides/hackathon17... · 2017. 2. 15. · Ivan Laptev ivan.laptev@inria.fr WILLOW, INRIA/ENS/CNRS, Paris Weakly supervised](https://reader033.vdocuments.us/reader033/viewer/2022060600/6053d5ffcbad1e388c6b1557/html5/thumbnails/14.jpg)
Motivation: labeling bounding boxes is tedious
![Page 15: Weakly supervised learning from images and videolaptev/share/slides/hackathon17... · 2017. 2. 15. · Ivan Laptev ivan.laptev@inria.fr WILLOW, INRIA/ENS/CNRS, Paris Weakly supervised](https://reader033.vdocuments.us/reader033/viewer/2022060600/6053d5ffcbad1e388c6b1557/html5/thumbnails/15.jpg)
Are bounding boxes needed for training CNNs?
Image-level labels: Bicycle, Person
![Page 16: Weakly supervised learning from images and videolaptev/share/slides/hackathon17... · 2017. 2. 15. · Ivan Laptev ivan.laptev@inria.fr WILLOW, INRIA/ENS/CNRS, Paris Weakly supervised](https://reader033.vdocuments.us/reader033/viewer/2022060600/6053d5ffcbad1e388c6b1557/html5/thumbnails/16.jpg)
Motivation: image-level labels are plentiful
“Beautiful red leaves in a back street of Freiburg”
[Kuznetsova et al., ACL 2013]
http://www.cs.stonybrook.edu/~pkuznetsova/imgcaption/captions1K.html
![Page 17: Weakly supervised learning from images and videolaptev/share/slides/hackathon17... · 2017. 2. 15. · Ivan Laptev ivan.laptev@inria.fr WILLOW, INRIA/ENS/CNRS, Paris Weakly supervised](https://reader033.vdocuments.us/reader033/viewer/2022060600/6053d5ffcbad1e388c6b1557/html5/thumbnails/17.jpg)
Motivation: image-level labels are plentiful
“Public bikes in Warsaw during night”
https://www.flickr.com/photos/jacek_kadaj/8776008002/in/photostream/
![Page 18: Weakly supervised learning from images and videolaptev/share/slides/hackathon17... · 2017. 2. 15. · Ivan Laptev ivan.laptev@inria.fr WILLOW, INRIA/ENS/CNRS, Paris Weakly supervised](https://reader033.vdocuments.us/reader033/viewer/2022060600/6053d5ffcbad1e388c6b1557/html5/thumbnails/18.jpg)
GoalTraining input
Test output
image-level labels:
Person
Chair
Airplane
+ Reading
Riding bike
Running
More details in http://www.di.ens.fr/willow/research/weakcnn/
… …
![Page 19: Weakly supervised learning from images and videolaptev/share/slides/hackathon17... · 2017. 2. 15. · Ivan Laptev ivan.laptev@inria.fr WILLOW, INRIA/ENS/CNRS, Paris Weakly supervised](https://reader033.vdocuments.us/reader033/viewer/2022060600/6053d5ffcbad1e388c6b1557/html5/thumbnails/19.jpg)
Approach: search over object’s location
at the training time
1. Fully convolutional network
2. Image-level aggregation (max-pool)
3. Multi-label loss function (allow multiple objects in image)
See also [Papandreou et al. ’15, Sermanet et al. ’14, Chaftield et al.’14]
Max-pool
over image
Per-image score
FCaFCbC1-C2-C3-C4-C5 FC6 FC7
4096-dimvector9216-
dimvector
4096-dimvector
…
motorbike
person
diningtable
pottedplant
chair
car
bus
train
…
Max
Oquab, Bottou, Laptev and Sivic CVPR 2015
![Page 20: Weakly supervised learning from images and videolaptev/share/slides/hackathon17... · 2017. 2. 15. · Ivan Laptev ivan.laptev@inria.fr WILLOW, INRIA/ENS/CNRS, Paris Weakly supervised](https://reader033.vdocuments.us/reader033/viewer/2022060600/6053d5ffcbad1e388c6b1557/html5/thumbnails/20.jpg)
Training
Motorbikes
Evolution of
localization
score maps
over training
epochs
![Page 21: Weakly supervised learning from images and videolaptev/share/slides/hackathon17... · 2017. 2. 15. · Ivan Laptev ivan.laptev@inria.fr WILLOW, INRIA/ENS/CNRS, Paris Weakly supervised](https://reader033.vdocuments.us/reader033/viewer/2022060600/6053d5ffcbad1e388c6b1557/html5/thumbnails/21.jpg)
Test results on 80 classes in Microsoft COCO dataset
![Page 22: Weakly supervised learning from images and videolaptev/share/slides/hackathon17... · 2017. 2. 15. · Ivan Laptev ivan.laptev@inria.fr WILLOW, INRIA/ENS/CNRS, Paris Weakly supervised](https://reader033.vdocuments.us/reader033/viewer/2022060600/6053d5ffcbad1e388c6b1557/html5/thumbnails/22.jpg)
Test results on 80 classes in Microsoft COCO dataset
![Page 23: Weakly supervised learning from images and videolaptev/share/slides/hackathon17... · 2017. 2. 15. · Ivan Laptev ivan.laptev@inria.fr WILLOW, INRIA/ENS/CNRS, Paris Weakly supervised](https://reader033.vdocuments.us/reader033/viewer/2022060600/6053d5ffcbad1e388c6b1557/html5/thumbnails/23.jpg)
Test results on 80 classes in Microsoft COCO dataset
![Page 24: Weakly supervised learning from images and videolaptev/share/slides/hackathon17... · 2017. 2. 15. · Ivan Laptev ivan.laptev@inria.fr WILLOW, INRIA/ENS/CNRS, Paris Weakly supervised](https://reader033.vdocuments.us/reader033/viewer/2022060600/6053d5ffcbad1e388c6b1557/html5/thumbnails/24.jpg)
Test results on 80 classes in Microsoft COCO dataset
![Page 25: Weakly supervised learning from images and videolaptev/share/slides/hackathon17... · 2017. 2. 15. · Ivan Laptev ivan.laptev@inria.fr WILLOW, INRIA/ENS/CNRS, Paris Weakly supervised](https://reader033.vdocuments.us/reader033/viewer/2022060600/6053d5ffcbad1e388c6b1557/html5/thumbnails/25.jpg)
Test results on 80 classes in Microsoft COCO dataset
![Page 26: Weakly supervised learning from images and videolaptev/share/slides/hackathon17... · 2017. 2. 15. · Ivan Laptev ivan.laptev@inria.fr WILLOW, INRIA/ENS/CNRS, Paris Weakly supervised](https://reader033.vdocuments.us/reader033/viewer/2022060600/6053d5ffcbad1e388c6b1557/html5/thumbnails/26.jpg)
Results for weakly-supervised
action recognition
in Pascal VOC’12 dataset
![Page 27: Weakly supervised learning from images and videolaptev/share/slides/hackathon17... · 2017. 2. 15. · Ivan Laptev ivan.laptev@inria.fr WILLOW, INRIA/ENS/CNRS, Paris Weakly supervised](https://reader033.vdocuments.us/reader033/viewer/2022060600/6053d5ffcbad1e388c6b1557/html5/thumbnails/27.jpg)
Test results for 10 action classes in Pascal VOC12
![Page 28: Weakly supervised learning from images and videolaptev/share/slides/hackathon17... · 2017. 2. 15. · Ivan Laptev ivan.laptev@inria.fr WILLOW, INRIA/ENS/CNRS, Paris Weakly supervised](https://reader033.vdocuments.us/reader033/viewer/2022060600/6053d5ffcbad1e388c6b1557/html5/thumbnails/28.jpg)
Test results for 10 action classes in Pascal VOC12
![Page 29: Weakly supervised learning from images and videolaptev/share/slides/hackathon17... · 2017. 2. 15. · Ivan Laptev ivan.laptev@inria.fr WILLOW, INRIA/ENS/CNRS, Paris Weakly supervised](https://reader033.vdocuments.us/reader033/viewer/2022060600/6053d5ffcbad1e388c6b1557/html5/thumbnails/29.jpg)
Test results for 10 action classes in Pascal VOC12
![Page 30: Weakly supervised learning from images and videolaptev/share/slides/hackathon17... · 2017. 2. 15. · Ivan Laptev ivan.laptev@inria.fr WILLOW, INRIA/ENS/CNRS, Paris Weakly supervised](https://reader033.vdocuments.us/reader033/viewer/2022060600/6053d5ffcbad1e388c6b1557/html5/thumbnails/30.jpg)
Test results for 10 action classes in Pascal VOC12
Failure cases
![Page 31: Weakly supervised learning from images and videolaptev/share/slides/hackathon17... · 2017. 2. 15. · Ivan Laptev ivan.laptev@inria.fr WILLOW, INRIA/ENS/CNRS, Paris Weakly supervised](https://reader033.vdocuments.us/reader033/viewer/2022060600/6053d5ffcbad1e388c6b1557/html5/thumbnails/31.jpg)
Weakly-supervised learning of
actions in video
from scripts and narrations
![Page 32: Weakly supervised learning from images and videolaptev/share/slides/hackathon17... · 2017. 2. 15. · Ivan Laptev ivan.laptev@inria.fr WILLOW, INRIA/ENS/CNRS, Paris Weakly supervised](https://reader033.vdocuments.us/reader033/viewer/2022060600/6053d5ffcbad1e388c6b1557/html5/thumbnails/32.jpg)
As the headwaiter takes them to a table they pass by the piano, and the woman looks at Sam. Sam, with a conscious effort, keeps his eyes on the keyboard as they go past. The headwaiter seats Ilsa...
34
![Page 33: Weakly supervised learning from images and videolaptev/share/slides/hackathon17... · 2017. 2. 15. · Ivan Laptev ivan.laptev@inria.fr WILLOW, INRIA/ENS/CNRS, Paris Weakly supervised](https://reader033.vdocuments.us/reader033/viewer/2022060600/6053d5ffcbad1e388c6b1557/html5/thumbnails/33.jpg)
As the headwaiter takes them to a table they pass by the piano, and the woman looks at Sam.Sam, with a conscious effort, keeps his eyes on the keyboard as they go past. The headwaiter seats Ilsa...
35
![Page 34: Weakly supervised learning from images and videolaptev/share/slides/hackathon17... · 2017. 2. 15. · Ivan Laptev ivan.laptev@inria.fr WILLOW, INRIA/ENS/CNRS, Paris Weakly supervised](https://reader033.vdocuments.us/reader033/viewer/2022060600/6053d5ffcbad1e388c6b1557/html5/thumbnails/34.jpg)
As the headwaiter takes them to a table they pass by the piano, and the woman looks at Sam. Sam, with a conscious effort, keeps his eyes on the keyboard as they go past. The headwaiter seats Ilsa...
36
![Page 35: Weakly supervised learning from images and videolaptev/share/slides/hackathon17... · 2017. 2. 15. · Ivan Laptev ivan.laptev@inria.fr WILLOW, INRIA/ENS/CNRS, Paris Weakly supervised](https://reader033.vdocuments.us/reader033/viewer/2022060600/6053d5ffcbad1e388c6b1557/html5/thumbnails/35.jpg)
As the headwaiter takes them to a table they pass by the piano, and the woman looks at Sam. Sam, with a conscious effort, keeps his eyes on the keyboard as they go past. The headwaiter seats Ilsa...
37
![Page 36: Weakly supervised learning from images and videolaptev/share/slides/hackathon17... · 2017. 2. 15. · Ivan Laptev ivan.laptev@inria.fr WILLOW, INRIA/ENS/CNRS, Paris Weakly supervised](https://reader033.vdocuments.us/reader033/viewer/2022060600/6053d5ffcbad1e388c6b1557/html5/thumbnails/36.jpg)
…
1172
01:20:17,240 --> 01:20:20,437
Why weren't you honest with me?
Why'd you keep your marriage a secret?
1173
01:20:20,640 --> 01:20:23,598
lt wasn't my secret, Richard.
Victor wanted it that way.
1174
01:20:23,800 --> 01:20:26,189
Not even our closest friends
knew about our marriage.
…
…
RICK
Why weren't you honest with me? Why
did you keep your marriage a secret?
Rick sits down with Ilsa.
ILSA
Oh, it wasn't my secret, Richard.
Victor wanted it that way. Not even
our closest friends knew about our
marriage.
…
01:20:17
01:20:23
subtitles movie script
• Scripts available for >500 movies (no time synchronization)
www.dailyscript.com, www.movie-page.com, www.weeklyscript.com …
• Subtitles (with time info.) are available for the most of movies
• Can transfer time to scripts by text alignment
Script-based video annotation
[Laptev, Marszałek, Schmid, Rozenfeld 2008]
![Page 37: Weakly supervised learning from images and videolaptev/share/slides/hackathon17... · 2017. 2. 15. · Ivan Laptev ivan.laptev@inria.fr WILLOW, INRIA/ENS/CNRS, Paris Weakly supervised](https://reader033.vdocuments.us/reader033/viewer/2022060600/6053d5ffcbad1e388c6b1557/html5/thumbnails/37.jpg)
Joint Learning of Actors and Actions
Rick? Rick?
Walks?Walks?
[Bojanowski et al. ICCV 2013]
Rick walks up behind Ilsa
![Page 38: Weakly supervised learning from images and videolaptev/share/slides/hackathon17... · 2017. 2. 15. · Ivan Laptev ivan.laptev@inria.fr WILLOW, INRIA/ENS/CNRS, Paris Weakly supervised](https://reader033.vdocuments.us/reader033/viewer/2022060600/6053d5ffcbad1e388c6b1557/html5/thumbnails/38.jpg)
Rick
Walks
Rick walks up behind Ilsa
Joint Learning of Actors and Actions[Bojanowski et al. ICCV 2013]
![Page 39: Weakly supervised learning from images and videolaptev/share/slides/hackathon17... · 2017. 2. 15. · Ivan Laptev ivan.laptev@inria.fr WILLOW, INRIA/ENS/CNRS, Paris Weakly supervised](https://reader033.vdocuments.us/reader033/viewer/2022060600/6053d5ffcbad1e388c6b1557/html5/thumbnails/39.jpg)
Formulation: Cost function
Rick
Ilsa
Sam
Actor labels Actor image features
Actor classifier
![Page 40: Weakly supervised learning from images and videolaptev/share/slides/hackathon17... · 2017. 2. 15. · Ivan Laptev ivan.laptev@inria.fr WILLOW, INRIA/ENS/CNRS, Paris Weakly supervised](https://reader033.vdocuments.us/reader033/viewer/2022060600/6053d5ffcbad1e388c6b1557/html5/thumbnails/40.jpg)
Formulation: Cost function
Person p appears at
least once in clip N :
p = Rick
Weak supervision
from scripts:
![Page 41: Weakly supervised learning from images and videolaptev/share/slides/hackathon17... · 2017. 2. 15. · Ivan Laptev ivan.laptev@inria.fr WILLOW, INRIA/ENS/CNRS, Paris Weakly supervised](https://reader033.vdocuments.us/reader033/viewer/2022060600/6053d5ffcbad1e388c6b1557/html5/thumbnails/41.jpg)
![Page 42: Weakly supervised learning from images and videolaptev/share/slides/hackathon17... · 2017. 2. 15. · Ivan Laptev ivan.laptev@inria.fr WILLOW, INRIA/ENS/CNRS, Paris Weakly supervised](https://reader033.vdocuments.us/reader033/viewer/2022060600/6053d5ffcbad1e388c6b1557/html5/thumbnails/42.jpg)
![Page 43: Weakly supervised learning from images and videolaptev/share/slides/hackathon17... · 2017. 2. 15. · Ivan Laptev ivan.laptev@inria.fr WILLOW, INRIA/ENS/CNRS, Paris Weakly supervised](https://reader033.vdocuments.us/reader033/viewer/2022060600/6053d5ffcbad1e388c6b1557/html5/thumbnails/43.jpg)
![Page 44: Weakly supervised learning from images and videolaptev/share/slides/hackathon17... · 2017. 2. 15. · Ivan Laptev ivan.laptev@inria.fr WILLOW, INRIA/ENS/CNRS, Paris Weakly supervised](https://reader033.vdocuments.us/reader033/viewer/2022060600/6053d5ffcbad1e388c6b1557/html5/thumbnails/44.jpg)
![Page 45: Weakly supervised learning from images and videolaptev/share/slides/hackathon17... · 2017. 2. 15. · Ivan Laptev ivan.laptev@inria.fr WILLOW, INRIA/ENS/CNRS, Paris Weakly supervised](https://reader033.vdocuments.us/reader033/viewer/2022060600/6053d5ffcbad1e388c6b1557/html5/thumbnails/45.jpg)
All problems solved?
![Page 46: Weakly supervised learning from images and videolaptev/share/slides/hackathon17... · 2017. 2. 15. · Ivan Laptev ivan.laptev@inria.fr WILLOW, INRIA/ENS/CNRS, Paris Weakly supervised](https://reader033.vdocuments.us/reader033/viewer/2022060600/6053d5ffcbad1e388c6b1557/html5/thumbnails/46.jpg)
![Page 47: Weakly supervised learning from images and videolaptev/share/slides/hackathon17... · 2017. 2. 15. · Ivan Laptev ivan.laptev@inria.fr WILLOW, INRIA/ENS/CNRS, Paris Weakly supervised](https://reader033.vdocuments.us/reader033/viewer/2022060600/6053d5ffcbad1e388c6b1557/html5/thumbnails/47.jpg)
Current solution: learn person-throws-cat-into-trash-bin classifier
Source: http://www.youtube.com/watch?v=eYdUZdan5i8
![Page 48: Weakly supervised learning from images and videolaptev/share/slides/hackathon17... · 2017. 2. 15. · Ivan Laptev ivan.laptev@inria.fr WILLOW, INRIA/ENS/CNRS, Paris Weakly supervised](https://reader033.vdocuments.us/reader033/viewer/2022060600/6053d5ffcbad1e388c6b1557/html5/thumbnails/48.jpg)
What is intention of this person?Is this scene dangerous?What is unusual in this scene?
Limitations of Current Methods
What is intention of this person?Is this scene dangerous?What is unusual in this scene?