using grammars for action recognition · aniket bera. video analysis with cfgs the “inverse...

20
Using Grammars for Action Recognition Aniket Bera

Upload: others

Post on 05-Aug-2020

4 views

Category:

Documents


0 download

TRANSCRIPT

Page 1: Using Grammars for Action Recognition · Aniket Bera. Video analysis with CFGs The “Inverse Hollywood problem”: From video to scripts and storyboards via causal analysis. Brand

Using Grammars for Action Recognition

Aniket Bera

Page 2: Using Grammars for Action Recognition · Aniket Bera. Video analysis with CFGs The “Inverse Hollywood problem”: From video to scripts and storyboards via causal analysis. Brand

Video analysis with CFGs

The “Inverse Hollywood problem”:

From video to scripts and storyboards via causal analysis.

Brand 1997

Action Recognition using Probabilistic Parsing.Bobick and Ivanov 1998

Recognizing Multitasked Activities from Video using

Stochastic Context-Free Grammar.

Moore and Essa 2001

13

Page 3: Using Grammars for Action Recognition · Aniket Bera. Video analysis with CFGs The “Inverse Hollywood problem”: From video to scripts and storyboards via causal analysis. Brand

CFG for human activities

enter detach leave enter detach attach touch touch detach attach leave

M. Brand. The "Inverse Hollywood Problem":From video to scripts and storyboards

via causal analysis. AAAI 1997.

14

Page 4: Using Grammars for Action Recognition · Aniket Bera. Video analysis with CFGs The “Inverse Hollywood problem”: From video to scripts and storyboards via causal analysis. Brand

Parse treeSCENE (Open up a PC)

IN ACTION (Open PC)

OUT IN

ADD ADD

enter detach leave enter

ACTION (unscrew) OUT

MOVE REMOVE

MOTION MOTION

detach attach touch touch detach attach leave

• Deterministic low-level primitive detection• Deterministic parsing

M. Brand. The "Inverse Hollywood Problem": From video to scripts and storyboards via causal analysis. AAAI 1997.

15

Page 5: Using Grammars for Action Recognition · Aniket Bera. Video analysis with CFGs The “Inverse Hollywood problem”: From video to scripts and storyboards via causal analysis. Brand

Stochastic CFGs

Action Recognition using Probabilistic Parsing.Bobick and Ivanov 1998

16

Page 6: Using Grammars for Action Recognition · Aniket Bera. Video analysis with CFGs The “Inverse Hollywood problem”: From video to scripts and storyboards via causal analysis. Brand

Gesture analysis with CFGs

Primitive recognition with HMMs

Action Recognition using Probabilistic Parsing. Bobick and Ivanov 1998 17

Page 7: Using Grammars for Action Recognition · Aniket Bera. Video analysis with CFGs The “Inverse Hollywood problem”: From video to scripts and storyboards via causal analysis. Brand

left-right

Action Recognition using Probabilistic Parsing. Bobick and Ivanov 1998 18

Page 8: Using Grammars for Action Recognition · Aniket Bera. Video analysis with CFGs The “Inverse Hollywood problem”: From video to scripts and storyboards via causal analysis. Brand

up-down

Action Recognition using Probabilistic Parsing. Bobick and Ivanov 1998 19

Page 9: Using Grammars for Action Recognition · Aniket Bera. Video analysis with CFGs The “Inverse Hollywood problem”: From video to scripts and storyboards via causal analysis. Brand

right-left

Action Recognition using Probabilistic Parsing. Bobick and Ivanov 1998 20

Page 10: Using Grammars for Action Recognition · Aniket Bera. Video analysis with CFGs The “Inverse Hollywood problem”: From video to scripts and storyboards via causal analysis. Brand

down-up

Action Recognition using Probabilistic Parsing. Bobick and Ivanov 1998 21

Page 11: Using Grammars for Action Recognition · Aniket Bera. Video analysis with CFGs The “Inverse Hollywood problem”: From video to scripts and storyboards via causal analysis. Brand

Parse Tree

S

RH

TOP UD BOT DU

LR RL

left-right up-down right-left down-up

22

Page 12: Using Grammars for Action Recognition · Aniket Bera. Video analysis with CFGs The “Inverse Hollywood problem”: From video to scripts and storyboards via causal analysis. Brand

Errors

Likelihood value over time (not discrete symbols)

HMM a

HMM b

Errors are inevitable…

but the grammar acts as a top-down constraint

Action Recognition using Probabilistic Parsing. Bobick and Ivanov 1998 23

Page 13: Using Grammars for Action Recognition · Aniket Bera. Video analysis with CFGs The “Inverse Hollywood problem”: From video to scripts and storyboards via causal analysis. Brand

Dealing with uncertainty & errors

Stolcke-Early (probabilistic) parser

SKIP rules to deal with insertion errors

HMM a

HMM b

HMM c

Action Recognition using Probabilistic Parsing. Bobick and Ivanov 1998 24

Page 14: Using Grammars for Action Recognition · Aniket Bera. Video analysis with CFGs The “Inverse Hollywood problem”: From video to scripts and storyboards via causal analysis. Brand

SCFG for Blackjack

Recognizing Multitasked Activities from Video usingStochastic Context-Free Grammar.

Moore and Essa 2001

• Deals with more complex activities• Deals with more error types

25

Page 15: Using Grammars for Action Recognition · Aniket Bera. Video analysis with CFGs The “Inverse Hollywood problem”: From video to scripts and storyboards via causal analysis. Brand

Stochastic Grammars: Overview

• Representation: Stochastic grammar• Terminals: object interactions• Context-sensitive due to internal scene models

• Domain: Towers of Hanoi• Requires activities with

strong temporal constraints

• Contributions• Showed recognition &

decomposition with veryweak appearance models

• Demonstrated usefulnessof feedback from high tolow-level reasoning components

Page 16: Using Grammars for Action Recognition · Aniket Bera. Video analysis with CFGs The “Inverse Hollywood problem”: From video to scripts and storyboards via causal analysis. Brand

Expectation Grammars(CVPR 2003)

• Analyze video of a person physically solving the Towers of Hanoi task

• Recognize valid activity

• Identify each move

• Segment objects

• Detect distracters / noise

Page 17: Using Grammars for Action Recognition · Aniket Bera. Video analysis with CFGs The “Inverse Hollywood problem”: From video to scripts and storyboards via causal analysis. Brand

System Overview

Page 18: Using Grammars for Action Recognition · Aniket Bera. Video analysis with CFGs The “Inverse Hollywood problem”: From video to scripts and storyboards via causal analysis. Brand

ToH: Low-Level Vision

Raw VideoBackground

Model

ForegroundComponents

Foreground andshadow detection

Page 19: Using Grammars for Action Recognition · Aniket Bera. Video analysis with CFGs The “Inverse Hollywood problem”: From video to scripts and storyboards via causal analysis. Brand

Low-Level Features• Explanation-based symbols

• Blob interaction events

• merge, split, enter, exit, tracked, noise

• Future Work: hidden, revealed, blob-part, coalesce

• All possible explanations generated• Inconsistent explanations heuristically pruned

Enter

Merge

Page 20: Using Grammars for Action Recognition · Aniket Bera. Video analysis with CFGs The “Inverse Hollywood problem”: From video to scripts and storyboards via causal analysis. Brand

Contributions

• Showed activity recognition and decomposition without appearance models

• Demonstrated usefulness of feedback from high-level, long-term interpretations to low-level, short-term decisions