on detection of multiple object instances using hough transforms
DESCRIPTION
On Detection of Multiple Object Instances using Hough Transforms. Olga Barinova Moscow State University. Victor Lempitsky University of Oxford. Pushmeet Kohli Microsoft Research Cambridge. Hough transforms. Object detection → peaks identification in Hough images Beyond lines!!! - PowerPoint PPT PresentationTRANSCRIPT
On Detection of Multiple Object Instances using Hough Transforms
Olga BarinovaMoscow State
University
Victor LempitskyUniversity of Oxford
Pushmeet Kohli Microsoft Research
Cambridge
On Detection of Multiple Object Instances Using Hough Transforms
Hough transforms
o Object detection → peaks identification in Hough images
o Beyond lines!!! Ballard 1983 – Other primitives Lowe, ICCV 1999 – Object detection Leibe, Schiele BMVC 2003 – Object class detection Last CVPR: Maji& Malik, Gall& Lempitsky, Gu et al. …
On Detection of Multiple Object Instances Using Hough Transforms
Example from Gall & Lempitsky CVPR 2009
Category-level object detection
On Detection of Multiple Object Instances Using Hough Transforms
Category-level object detection
?
On Detection of Multiple Object Instances Using Hough Transforms
Multiple lines detection
o Identifying the peaks in Hough images is highly nontrivial in case of multiple close objects
o Postprocessing (e.g. non-maximum suppression) is usually used
o Our framework is similar to Hough Transforms but doesn’t require finding local maxima and suppresses non-maxima automatically
On Detection of Multiple Object Instances Using Hough Transforms
Our frameworkHough spaceElements space
Voting elements
Hypotheses
On Detection of Multiple Object Instances Using Hough Transforms
Our framework
1
2
3
y – labelling of hypotheses, binary variables:
1 = object is present,
0 = otherwise
Hough spaceElements space
x – labelling of voting elements,
xi = index of hypothesis,
if element votes for hypothesis,
xi = 0, if element votes for background
On Detection of Multiple Object Instances Using Hough Transforms
Our framework
1
2
3x – labelling of voting
elements,
xi = index of hypothesis,
if element votes for hypothesis,
xi = 0, if element votes for background
x2=1x3=1
x4=2
x5=2x6=2
x7=0
x8=2
x1=1
y1=1 y2 =1
y3=0
Key idea : joint MAP-inference in x and y
Hough spaceElements space
y – labelling of hypotheses, binary variables:
1 = object is present,
0 = otherwise
On Detection of Multiple Object Instances Using Hough Transforms
Probabilistic derivation
Likelihood Term
o Assume that given the existing objects y and the hypotheses assignments x, the distributions of the descriptors of voting elements are independent
o Less crude than the independence assumption implicitly made by the traditional Hough
Prior Term
o Occam razor (or MDL) prior penalizes the number of the active hypotheses
On Detection of Multiple Object Instances Using Hough Transforms
Probabilistic derivation123
0
123
0
123
0
123
0
123
0
10
10
10
1x
2x
3x
4x
5x
1y
2y
3y
hypotheses
how likely is that voting element belongs to an object
Corresponds to the votes in standard Hough transform: Training stays the same!
“MDL” prior:
λ, if yh = 1
0, otherwise
-∞ if xi = h, and yh = 0
0, otherwise
voting elements
Problem known as facility location[Delong et al. CVPR 2010] (today’s poster)
looks at facility location with wider set of priors
On Detection of Multiple Object Instances Using Hough Transforms
Probabilistic derivation123
0
123
0
123
0
123
0
123
0
10
10
10
1x
2x
3x
4x
5x
1y
2y
3y
hypotheses
voting elements
o Tried different methods for MAP-inference belief propagation simulated annealing
o They work well but don’t allow using large number of hypotheses graph becomes huge and dense sparsification heuristics required
On Detection of Multiple Object Instances Using Hough Transforms
Probabilistic derivation123
0
123
0
123
0
123
0
123
0
10
10
10
1x
2x
3x
4x
5x
1y
2y
3y
voting elements
hypotheses
o If labeling of y is given the values of xi are independent
o After maximizing out x we get:
o Large-clique, submodular o Greedy algorithm is as good as anything else
(in terms of the approximation factor)
o Greedy inference ~ iterative Hough voting
On Detection of Multiple Object Instances Using Hough Transforms
Greedy maximization for our energy
Greedily add detections starting from the empty set
For each iteration
1. do the voting:
o Set h0 = the overall maximum of HoughImage
3. If HoughImage(h0) > λ, add h0 to detection set, else terminate
Sum over all voting elements
Maximum over Hough votes for the hypotheses g that have already been switched on, including ‘background’
“standard” Hough vote for element i
On Detection of Multiple Object Instances Using Hough Transforms
Inference
Using the Hough forest trained in [Gall&Lempitsky CVPR09]
Datasets from [Andriluka et al. CVPR 2008](with strongly occluded pedestrians added)
On Detection of Multiple Object Instances Using Hough Transforms
Results for pedestrians detection
White = correct detectionGreen = missing objectRed = false positive
Our frameworkHough transform + non-maximum suppression
On Detection of Multiple Object Instances Using Hough Transforms
Results for pedestrians detection
Blue = Hough transform + non-maximum suppressionLight-blue = our framework
Precision
Recall
Precision
Recall
TUD-crossingTUD-campus
On Detection of Multiple Object Instances Using Hough Transforms
Results for lines detection
Our framework
Hough + NMS
York Urban DB, Elder&Estrada ECCV 2008
o our framework is able to discern very close yet distinct lines, and is in general much less plagued by spurious detections
On Detection of Multiple Object Instances Using Hough Transforms
Conclusion
o Framework for detecting multiple objects, greedy inference ~ iterated Hough transform
o No need to find local maxima and suppress non-maxima – just take the only global maximum
o Probabilistic model allows for extensions (ECCV paper coming: lines + vanishing points + horizon + zenith)
o Training stays the same as for the recent Hough-based framework
o Code available at the project page: http://graphics.cs.msu.ru/en/science/research/machinelearning/hough
Thank you for your attention!