Slide 1

Slide 1 text

Towards Realistic Predictors Pei Wang and Nuno Vasconcelos Statistical and Visual Computing Lab, UC San Diego mlKitchen X 2018.11.29 @_lunardog_ http://openaccess.thecvf.com/content_ECCV_2018/papers/Pei_Wang_Towards_Realistic_Predictors_ECCV_2018_paper.pdf

Slide 2

Slide 2 text

Self-Introduction ● Call me Leszek ● Born in Poland ● Living in Japan since 2010 ● Cookpad R&D since 2016 ● I consume too much science fiction ● I’m bad at selfies

Slide 3

Slide 3 text

No content

Slide 4

Slide 4 text

No content

Slide 5

Slide 5 text

No content

Slide 6

Slide 6 text

No content

Slide 7

Slide 7 text

Towards Realistic Predictors Pei Wang and Nuno Vasconcelos Statistical and Visual Computing Lab, UC San Diego http://openaccess.thecvf.com/content_ECCV_2018/papers/Pei_Wang_Towards_Realistic_Predictors_ECCV_2018_paper.pdf

Slide 8

Slide 8 text

Define “Realistic” optimistic pessimistic realistic

Slide 9

Slide 9 text

https://snappygoat.com/

Slide 10

Slide 10 text

https://snappygoat.com/

Slide 11

Slide 11 text

https://www.abc15.com

Slide 12

Slide 12 text

A more benign example...

Slide 13

Slide 13 text

Classifier Food / non-food classifier food plant person pet … other

Slide 14

Slide 14 text

No content

Slide 15

Slide 15 text

No content

Slide 16

Slide 16 text

No content

Slide 17

Slide 17 text

No content

Slide 18

Slide 18 text

No content

Slide 19

Slide 19 text

No content

Slide 20

Slide 20 text

Let’s start with what we can easily classify

Slide 21

Slide 21 text

No content

Slide 22

Slide 22 text

No content

Slide 23

Slide 23 text

No content

Slide 24

Slide 24 text

What is “hard” anyway?

Slide 25

Slide 25 text

HP-Net = Hardness Predictor HP-Net hardness

Slide 26

Slide 26 text

Classifier HP-Net Adversarial training with a hardness predictor

Slide 27

Slide 27 text

Hardness Predictor Loss bi y s e t y s u w r e s ma c mi zi t K l a k-Le b di g e b en t di r i n d a m i m n = 1 − p c

Slide 28

Slide 28 text

Classifier Loss weighted by hardness we t ro -en p ma h er p e (la r ) mo po n , w i as e m s (lo s) ar en s or c

Slide 29

Slide 29 text

Classifier HP-Net Training 1. train classifier F and HP-Net S jointly on training set D 2. run S on D and eliminate hard examples, to create realistic training set D′ 3. learn realistic classifier F′ on D′, with S fixed 4. output pair S, F′ 5. GOTO 1 D F S

Slide 30

Slide 30 text

Can’t we just use confidence scores?

Slide 31

Slide 31 text

No content

Slide 32

Slide 32 text

Hardness progression during training

Slide 33

Slide 33 text

Do we need two separate models?

Slide 34

Slide 34 text

Classifier + HP-Net +

Slide 35

Slide 35 text

No content

Slide 36

Slide 36 text

Do we need to fine tune?

Slide 37

Slide 37 text

C - normal classifier F - realistic predictor without fine-tuning (just rejection) F’ - realistic predictor, fine-tuned on samples accepted by HP-Net

Slide 38

Slide 38 text

No content

Slide 39

Slide 39 text

Conclusions ● There are times when it’s OK to skip hard samples ● ...and times when it’s BEST to reject hard samples ● The paper introduces a GAN-like architecture to train any classifier with its own hardness predictor ● Training with a hardness predictor improves accuracy ● HP-Net should be trained jointly with the classifier, but in an alternating order ● HP-Net solves a different problem from the Classifier, should be a separate model ● ….but best results are when the architectures are the same

Slide 40

Slide 40 text

https://ja.wikipedia.org/wiki/2001年宇宙の旅    うちゅうのたび 2001年宇宙の旅 『2001年宇宙の旅』(にせんいちねんう ちゅうのたび、原題:2001: A Space Odyssey)は、アーサー・C・クラークとスタン リー・キューブリックのアイデアをまとめた ストーリーに基いて製作された、SF映画お よびSF小説である。 2001: A Space Odyssey is a 1968 epic science fiction film produced and directed by Stanley Kubrick. The screenplay was written by Kubrick and Arthur C. Clarke, and was inspired by Clarke's short story "The Sentinel". A novel also called 2001: A Space Odyssey, written concurrently with the screenplay, was published soon after the film was released. 2001: A Space Odyssey https://en.wikipedia.org/wiki/2001:_A_Space_Odyssey_(film)

Slide 41

Slide 41 text

It’s the movie with the mysterious black block, and classical music in space.

Slide 42

Slide 42 text

Spaceships don’t make a “Whoosh!” sound, there’s classical music instead.

Slide 43

Slide 43 text

Fra Dav HA 9000 is r pe in f e n a om d i n p e h Frank and Dave don’t trust HAL. HAL is a realistic predictor and doesn’t always follow orders.

Slide 44

Slide 44 text

Open the pod bay doors, HAL!

Slide 45

Slide 45 text

I’m sorry, Dave. I’m afraid I can’t do that. Fra HAL kills Frank.

Slide 46

Slide 46 text

Dave has no choice but to deactivate HAL. Daisy… Daisy…

Slide 47

Slide 47 text

Dedicated to Douglas Rain (March 13, 1928 – November 11, 2018) known as the Voice of HAL