Larry Zitnick Facebook AI Research 1984 Neocognitron 1983 Recognition 1984 2016 Data GPUs Backprop Neocognitron 1983 AlexNet 2012 Recognition 1984 2016 Data GPUs Backprop Recognition ID: 585734
Download Presentation The PPT/PDF document "A Visual Stepping Stone to AI" is the property of its rightful owner. Permission is granted to download and print the materials on this web site for personal, non-commercial use only, and to display it on your personal computer provided you do not modify the materials and that you retain all copyright notices contained in the materials. By downloading content from our website, you accept the terms of this agreement.
Slide1Slide2
A Visual Stepping Stone to AI
Larry Zitnick
Facebook AI ResearchSlide3
1984
Neocognitron
, 1983
Recognition?Slide4
1984
2016
Data
GPUs
Backprop
Neocognitron
, 1983
AlexNet
, 2012
RecognitionSlide5
1984
2016
Data
GPUs
Backprop
Recognition
2048
AI Slide6
1984
2016
Data
GPUs
Backprop
Recognition
2048
AI
More data
More compute
BackpropSlide7
A man riding a wave on a surfboard in the water.
A giraffe standing in the grass next to a tree.
Mind’s Eye: A Recurrent Visual Representation for Image Caption Generation, Chen and Zitnick, CVPR 2015.Slide8
A man riding a
motorcycle on a beach.
An airplane is parked on the
tarmac at an airport.
Building Machines That Learn and Think Like People,
Lake et al., ArXiv 1604.00289, 2016Slide9
MIRRORSlide10
1984
2016
Data
GPUs
Backprop
Recognition
2048
AI
More Data
More Compute
BackpropSlide11
1984
2016
Data
GPUs
Backprop
Recognition
AI ? Slide12
LEARNINGSlide13
I didn’t look closely but I think there’s a cat.
There is a cat in this image.
I’m not telling you anything!!!
I’ll give you a $
1
if you find a cat image.
Supervised
Semi-supervised
Unsupervised
ReinforcementSlide14
SupervisedSlide15
Semi-supervised learning
A store display that has a lot of bananas on sale.
A
yellow
Vespa parked in a lot with other cars.Slide16
fence
pink
Unlikely
Likely
Learning Visual Classifiers using Human-centric Annotations Misra et al., CVPR 2016Slide17
Reinforcement learning
MazeBase
:
A sandbox for experimenting with reinforcement learning, reasoning, and planning.
MazeBase: A Sandbox for Learning From GamesSukhbaatar
et al., arXiv, 2016Slide18
Unsupervised learning
Correct
Wrong
Unsupervised Learning using Sequential Verification for Action Recognition, Misra et al.,
arXiv
, 2016Slide19
REASONING
Visual
^Slide20
Mary went to the hallway.
John moved to the bathroom.
Mary travelled to the kitchen.
Where is Mary? Dialog-based Language Learning, Weston, ArXiv 1604.06045, 2016
UNDERSTANDINGSlide21
Learning Physical Intuition of Block Towers by Example
Lerer
et al., arXiv 1603.01312, 2016
PREDICTINGSlide22
We Are Humor Beings: Understanding and Predicting Visual Humor,
Chandrasekaran et al., CVPR 2016
DATASlide23
Visual Question Answering
MEASURING
VQA: Visual Question Answering
Antol
et al., ICCV 2015
Is this a vegetarian pizza?
Does this person have 20/20 vision?Slide24
Looking forwardSlide25
1984
2016
Data
GPUs
Backprop
Recognition
2048
AI
Learning
ReasoningSlide26
1984
2016
Data
GPUs
Backprop
Recognition
Reasoning
2048
AI
LearningSlide27
1984
2016
Recognition
20**
AI
Learning
ReasoningSlide28
Facebook AI Research