ALE SOFTWARE-for academic use.
The Automatic Labelling
Environment
Ľubor Ladický, Philip H.S. Torr.
This is the Oxford Brookes Vision
Group code for semantic segmentation the code for which is available for
academic use, in order to access click HERE
to agree to the licence.
This code is based on a body of
work for scene understanding-recognizing objects and recovering their depths.
The work started off with a
method for providing an object label for each pixel in the scene based on our PN
model [IJCV 2009], the code allows you to train up such a classifier if you
provide your own training data, as well as some pre trained classifiers. The
work took best paper at ECCV 2010 as here we showed how to encode costs on
co-occurrence on sets of labels, something we have found to improve performance
as the number of classes increases.
We later extended this work to
also yield the number of objects in the scene [ECCV 2010] (providing a solution
to the “things” vs “stuff” problem, things are
objects with limited spatial extent such as cars, dogs etc. and stuff appears
more amorphous such as sky, vegetation etc.).
The work also allows us to make
a joint inference on depth labels and object labels [BMVC 2010 best paper] and
code for that is included, the data set for that work is here,
we added depth labels to the existing Leuven data set. A talk giving an
overview is here (caution 44 MB).
Publications
Ľubor Ladický,
Chris Russell, Pushmeet Kohli, Philip H.S. Torr
Graph Cut based
Inference with Co-occurrence Statistics
Proceedings of the Eleventh European Conference on Computer Vision,
2010.
Best Paper
Award
Ľubor Ladický,
Paul Sturgess, Karteek Alahari, Chris Russell, Philip H.S. Torr
What,Where & How Many? Combining
Object Detectors and CRFs
Proceedings of the Eleventh European Conference on Computer Vision,
2010.
Ľubor Ladický,
Paul Sturgess, Chris Russell, Sunando Sengupta, Yalin
Bastanlar, William Clocksin, Philip H.S. Torr
Joint Optimisation for
Object Class Segmentation and Dense Stereo Reconstruction
Proceedings British Machine Vision Conference, 2010.
Leuven Dataset
Best Paper
Award
Chris Russell, Ľubor Ladický, Pushmeet Kohli, Philip H.S. Torr
Exact and Approximate
Inference in Associative Hierarchical Networks using Graph Cuts
The 26th Conference on Uncertainty in Artificial Intelligence, 2010.
Pushmeet Kohli, Ľubor Ladický, Philip H.S. Torr
Robust Higher Order
Potentials for Enforcing Label Consistency
Proceedings of the International Journal of Computer Vision, 2009.
Ľubor Ladický, Chris Russell, Pushmeet Kohli,
Philip H.S. Torr
Associative Hierarchical
CRFs for Object Class Image Segmentation
Proceedings IEEE Twelfth International Conference on Computer Vision,
2009.
Paul Sturgess, Karteek Alahari, Ľubor
Ladický, Philip H.S. Torr
Combining Appearance and
Structure from Motion Features for Road Scene Understanding
Proceedings British Machine Vision Conference, 2009.
Pushmeet Kohli, Ľubor Ladický, Philip H.S. Torr
Robust Higher Order
Potentials for Enforcing Label Consistency
Proceedings IEEE Conference of Computer Vision and Pattern Recognition,
2008. Dataset (MSRC with accurate boundaries).
P. Kohli, M. Pawan Kumar and P.H.S. Torr. P3 & Beyond: Move Making Algorithms for Solving Higher Order Functions.
In IEEE Trans Pattern Analysis and Machine Intelligence, Volume 31, Issue 9, Pages 1645-1656 , 2008.
