Efficient Inference in Fully Connected CRFs with Gaussian ...
Associative Hierarchical CRFs for Object Class Image Segmentation Ľubor Ladický 1 1 Oxford Brookes...
-
Upload
shawn-hunt -
Category
Documents
-
view
225 -
download
1
Transcript of Associative Hierarchical CRFs for Object Class Image Segmentation Ľubor Ladický 1 1 Oxford Brookes...
Associative Hierarchical CRFs for Object Class Image
Segmentation
Ľubor Ladický1
1Oxford Brookes University 2Microsoft Research Cambridge
Based on the work done with :
Chris Russell1, Pushmeet Kohli1,2 , Paul Sturgess1 ,Karteek Alahari1 , Philip H.S. Torr1
Overview
Object-class Segmentation problem
Datasets
MSRC dataset
VOC2009 dataset
CamVid dataset
Standard approaches
Pairwise CRF over Pixels
Pairwise CRF over Segments
Detection-driven Segmentation
Model
Robust-PN model
Associative Hierarchical CRF
Detectors in AHCRF
Label co-occurrence in CRFs
Potential training
Inference
Results
Discussion
General overview Our approach
Overview
Object-class Segmentation problem
Datasets
MSRC dataset
VOC2009 dataset
CamVid dataset
Standard approaches
Pairwise CRF over Pixels
Pairwise CRF over Segments
Detection-driven Segmentation
Model
Robust-PN model
Associative Hierarchical CRF
Detectors in AHCRF
Label co-occurrence in CRFs
Potential training
Inference
Results
Discussion
General overview Our approach
Object-class Segmentation problem
MSRC Image Our Result
Aims to assign a label for each pixel of an image
Classifier trained on the training set
Performance evaluated on never seen test set
Overview
Object-class Segmentation problem
Datasets
MSRC dataset
VOC2009 dataset
CamVid dataset
Standard approaches
Pairwise CRF over Pixels
Pairwise CRF over Segments
Detection-driven Segmentation
Model
Robust-PN model
Associative Hierarchical CRF
Detectors in AHCRF
Label co-occurrence in CRFs
Potential training
Inference
Results
Discussion
General overview Our approach
Datasets
MSRC Dataset591 images 320 x 213
21 classes
MSRC Image MSRC Image MSRC ImageGround Truth Ground Truth Ground Truth
Datasets
VOC2009 Dataset1499 training + 750 test images 500 x 375
20 foreground classes + 1 background class
VOC Image VOC Image VOC ImageGround Truth Ground Truth Ground Truth
Datasets
CamVid Dataset367 training + 233 test video frames 960 x 720
11 classes
CamVid Image CamVid Image CamVid ImageGround Truth Ground Truth Ground Truth
Overview
Object-class Segmentation problem
Datasets
MSRC dataset
VOC2009 dataset
CamVid dataset
Standard approaches
Pairwise CRF over Pixels
Pairwise CRF over Segments
Detection-driven Segmentation
Model
Robust-PN model
Associative Hierarchical CRF
Detectors in AHCRF
Label co-occurrence in CRFs
Potential training
Inference
Results
Discussion
General overview Our approach
Pairwise CRF over Pixels
CRF
Shotton ECCV06
Pairwise CRF over Pixels
No quantization errors
Lacks long range interactions
Results oversmoothed
Better performance on CamVid & MSRC datasets
Pairwise CRF over Segments
Yang et al. CVPR07, Batra et al. CVPR08,
Shi, Malik PAMI2000, Comaniciu, Meer PAMI2002, Felzenschwalb, Huttenlocher, IJCV2004
Pairwise CRF over Segments
Allows long range interactions
Better performance for VOC dataset
Can not recover from incorrect segmentation
Impossible to obtain perfect unsupervised segmentation
Pairwise CRF over Segments
Allows long range interactions
Better performance for VOC dataset
Can not recover from incorrect segmentation
Impossible to obtain perfect unsupervised segmentation
Detection-driven Segmentation
Larlus et al. CVPR08, Felzenszwalb et al. CVPR08, Vedaldi et al. ICCV09
GrabCut
Detector
Detection-driven Segmentation
The best performance for VOC dataset
Can not be used for background classes
Can not recover from incorrect detection
Overview
Object-class Segmentation problem
Datasets
MSRC dataset
VOC2009 dataset
CamVid dataset
Standard approaches
Pairwise CRF over Pixels
Pairwise CRF over Segments
Detection-driven Segmentation
Model
Robust-PN model
Associative Hierarchical CRF
Detectors in AHCRF
Label co-occurrence in CRFs
Potential training
Inference
Results
Discussion
General overview Our approach
Robust PN approach
Kohli, Ladický, Torr CVPR08
Robust PN approach
Segment consistency as a weak constraint
Robust to misleading segmentations
Allows multiple overlapping segmentations
General formulation not used in application
Unary and pairwise potentials only at the pixel level
Robust PN reformulation
Ladický, Russell, Kohli, Torr ICCV09
Old formulation of Robust PN potential
is equivalent to pairwise formulation
where
Robust PN reformulation
Ladický, Russell, Kohli, Torr ICCV09
Associative Hierarchical CRF
Allows unary potentials for region variables
Allows pairwise potentials for region variables
Allows multiple layers and multiple hierarchies
Zhu et al. NIPS2008, Lim et al. ICCV2009, Hinton 2002
Analysis of the new model
Ladický, Russell, Kohli, Torr ICCV09
Let's have one segmentation and potentials only over segment level
Analysis of the new model
Let's have one segmentation and potentials only over segment level
Energy of two pixels from the same clique is symmetric and semi-metric
Minimum will be segment-consistent
The cost of every segment consistent labelling is the same as the cost of the pairwise CRF labelling over segments
Equivalent to pairwise CRF over segmentsLadický, Russell, Kohli, Torr ICCV09
Associative Hierarchical CRF
Merges information over multiple scales
Allows multiple hierarchies
Allows long range interactions
Easy to learn weights
Interlayer connection limited(?) to associative relationship
Ladický, Russell, Kohli, Torr ICCV09
Comparison to other methods
Ladický, Russell, Kohli, Torr ICCV09
Robust PN Segment CRF Hierarchical CRF
Detectors inAssociative Hierarchical CRFs
Ladický, Alahari, Sturgess, Russell, Torr CVPR10 (submitted)
GrabCutDetector
Detector potential takes the form
where
Detectors inAssociative Hierarchical CRFs
Ladický, Alahari, Sturgess, Russell, Torr CVPR10 (submitted)
Special 2-label case of Robust PN
Allows multiple overlapping detections
Can recover from incorrect detection
Same inference methods may be applied as for Associative Hierarchical CRFs
We can distinguish between different instances of objects in the final result
Label co-occurrence in CRFs
Ladický, Russell, Kohli, Torr CVPR10 (submitted)
Possible labelling ?
Why not ?
Label co-occurrence in CRFs
Ladický, Russell, Kohli, Torr CVPR10 (submitted)
Our requirements
Global Energy function – no hard decisions
Invariance to number of pixels taking given label
Efficiency – should be tractable
Standard approaches
Hard decisions in the preprocessing step (Csurka et al. BMVC08)
Unary potential (Torralba et al. ICCV03)
Pairwise potential (Rabinovich et al. ICCV07, CVPR08)
Only solution : EC = C(L(x))
Overview
Object-class Segmentation problem
Datasets
MSRC dataset
VOC2009 dataset
CamVid dataset
Standard approaches
Pairwise CRF over Pixels
Pairwise CRF over Segments
Detection-driven Segmentation
Model
Robust-PN model
Associative Hierarchical CRF
Detectors in AHCRF
Label co-occurrence in CRFs
Potential training
Inference
Results
Discussion
General overview Our approach
Multi-feature BoostedUnary Pixel Potential
Ladický, Russell, Kohli, Torr ICCV09Sturgess, Alahari, Ladický, Torr BMVC09
Unary likelihoods based on spatial configuration (Shotton et al. ECCV06)
Classifier trained using boosting
Multi-feature BoostedUnary Segment Potential
Ladický, Russell, Kohli, Torr ICCV09
Classifier trained using boosting
Other Potentials
Ladický, Russell, Kohli, Torr ICCV09Ladický, Russell, Kohli, Torr CVPR10 (submitted)
Ladický, Alahari, Sturgess, Russell, Torr CVPR10 (submitted)
• Intensity dependent pairwise pixel potential
• Colour EMD-distance dependent pairwise segment potential
• Detector potentials based on off-the-shelf detectors (Felzenszwalb et al. CVPR08, Vedaldi et al. ICCV09)
• Generatively trained Co-occurence potential
Overview
Object-class Segmentation problem
Datasets
MSRC dataset
VOC2009 dataset
CamVid dataset
Standard approaches
Pairwise CRF over Pixels
Pairwise CRF over Segments
Detection-driven Segmentation
Model
Robust-PN model
Associative Hierarchical CRF
Detectors in AHCRF
Label co-occurrence in CRFs
Potential training
Inference
Results
Discussion
General overview Our approach
Inference over Hierarchical CRF
Russell, Ladický, Kohli, Torr AISTATS10 (submitted)
Inference over Hierarchical CRF
Problem is NP-hard
Any message passing algorithm (TRW-S, BP, ..) or ICM can be applied to pairwise model
αβ-swap (potentials must be semi-metric)
Ishikawa construction over (α-F-β transition)
αexpansion (potentials must be metric)
Reparametrization of interlayer connection to metric potential
Ishikawa construction over (α-F-old transition)
For more details read our technical report
Russell, Ladický, Kohli, Torr AISTATS10 (submitted)
Inference over Hierarchical CRF
Problem is NP-hard
Any message passing algorithm (TRW-S, BP, ..), ICM or LP relaxation can be applied to pairwise model
αβ-swap (potentials must be semi-metric)
Ishikawa construction over (α-F-β transition)
αexpansion (potentials must be metric)
Reparametrization of interlayer connection to metric potential
Ishikawa construction over (α-F-old transition)
For more details read our technical report
Russell, Ladický, Kohli, Torr AISTATS10 (submitted)
Inference over Hierarchical CRF
Problem is NP-hard
Any message passing algorithm (TRW-S, BP, ..), ICM or LP relaxation can be applied to pairwise model
• αβ-swap (potentials must be semi-metric)
• Ishikawa construction over (α-F-β transition)
αexpansion (potentials must be metric)
Reparametrization of interlayer connection to metric potential
Ishikawa construction over (α-F-old transition)
For more details read our technical report
Russell, Ladický, Kohli, Torr AISTATS10 (submitted)
Inference over Hierarchical CRF
Problem is NP-hard
Any message passing algorithm (TRW-S, BP, ..), ICM or LP relaxation can be applied to pairwise model
• αβ-swap (potentials must be semi-metric)
• Ishikawa construction over (α-F-β transition)
• α-expansion (potentials must be metric)
• Reparametrization of interlayer connection to metric potential
• Ishikawa construction over (α-F-old transition)
Russell, Ladický, Kohli, Torr AISTATS10 (submitted)
Inference over Hierarchical CRF
Russell, Ladický, Kohli, Torr AISTATS10 (submitted)
Inference for Co-occurence
Integer programming formulation possible using one variable for each subset of the label set
LP relaxation applicable by relaxing IP program
• αβ-swap possible if the cooccurence cost C(L(x)) is monotonically increasing with respect to L(x)
• α-expansion approximate move possible if the cooccurence cost C(L(x)) is monotonically increasing with respect to L(x)
Ladický, Russell, Kohli, Torr CVPR10 (submitted)
Overview
Object-class Segmentation problem
Datasets
MSRC dataset
VOC2009 dataset
CamVid dataset
Standard approaches
Pairwise CRF over Pixels
Pairwise CRF over Segments
Detection-driven Segmentation
Model
Robust-PN model
Associative Hierarchical CRF
Detectors in AHCRF
Label co-occurrence in CRFs
Potential training
Inference
Results
Discussion
General overview Our approach
VOC2009 Results
VOC2009 Results
VOC2009 Results
MSRC Results
MSRC Results
Qualitative comparison of results without and with co-occurence
Ladický, Russell, Kohli, Torr CVPR10 (submitted)
MSRC Image MSRC ImageWithout CO With CO With COWithout CO
CamVid Results
Sturgess, Alahari Ladický, Torr BMVC09Ladický, Alahari, Sturgess, Russell, Torr CVPR10 (submitted)
Our Result
CamVid Image
GroundTruth
Brostow et al.
Qualitative comparisons
Quantitative comparison of methods on VOC2009 dataset
Quantitative comparison of methods on MSRC dataset
Quantitative comparison of methods on CamVid dataset
Take home message
Take home message
Use our model
Take home message
Use our model
• Use your favourite potentials
Take home message
Use our model
• Use your favourite potentials
• Use your friend's favourite potentials
Take home message
Use our model
• Use your favourite potentials
• Use your friend's favourite potentials
• Use your friend's friend's favourite potentials
Take home message
Use our model
• Use your favourite potentials
• Use your friend's favourite potentials
• Use your friend's friend's favourite potentials
• Vision solved
Take home message
Use our model
• Use your favourite potentials
• Use your friend's favourite potentials
• Use your friend's friend's favourite potentials
• Vision solved (..almost)
Thank you
Questions?