Professional Documents
Culture Documents
Title
Road surface and curbstones estimation using conditional random fields
1
4 Integrating LIDAR into Stereo for Fast and Improved Disparity Computation
10
11 Pedestrian detection based on codebook features(bag of words) and boosting
17
19
Active contours/Snakes
20
http://www.lelaps.de/papers/badino_wdv2007.pdf
http://www.cv.utcluj.ro/tl_files/cv/publications/ProcessingDenseStereoUsingElevationMaps.pdf
http://www.lelaps.de/papers/badino_icra11.pdf
http://www.lelaps.de/papers/badino_3dimpvt11.pdf
http://www.cvlibs.net/publications/Geiger2010ACCV.pdf
http://ieeexplore-1ieee-1org-1ieee.rb-han.de.bosch.com/stamp/stamp.jsp?tp=&arnumber=6385492&tag=1
http://www.mi.auckland.ac.nz/tech-reports/12_SHRK_ACCV.pdf
http://ieeexplore-1ieee-1org-1ieee.rb-han.de.bosch.com/stamp/stamp.jsp?tp=&arnumber=5284778
http://users.soe.ucsc.edu/~manduchi/papers/IVS00.pdf
http://users.soe.ucsc.edu/~manduchi/papers/AutRobots.pdf
http://www.cogsys.cs.uni-tuebingen.de/publikationen/2011/khan2011civts.pdf
http://www-robotics.jpl.nasa.gov/publications/Arturo_Rankin/spie2009-rankin-7332-32-final.pdf
(none)
http://ieeexplore-1ieee-1org-1ieee.rb-han.de.bosch.com/stamp/stamp.jsp?tp=&arnumber=6646092&tag=1
http://ieeexplore-1ieee-1org-1ieee.rb-han.de.bosch.com/stamp/stamp.jsp?arnumber=05284804
http://rd.springer.com.springer.rb-han.de.bosch.com/chapter/10.1007%2F11556985_70
http://infoscience.epfl.ch/record/177415/files/Superpixel_PAMI2011-2.pdf
http://www.gcpr2013.uni-saarland.de/material/urtasun_flow.pdf
http://ttic.uchicago.edu/~rurtasun/publications/yamaguchi_et_al_cvpr13.pdf
http://ieeexplore-1ieee-1org-1ieee.rb-han.de.bosch.com/stamp/stamp.jsp?tp=&arnumber=6163004
http://infoscience.epfl.ch/record/177415/files/Superpixel_PAMI2011-2.pdf
http://www.gcpr2013.uni-saarland.de/material/urtasun_flow.pdf
http://ttic.uchicago.edu/~rurtasun/publications/yamaguchi_et_al_cvpr13.pdf
http://ieeexplore-1ieee-1org-1ieee.rb-han.de.bosch.com/stamp/stamp.jsp?tp=&arnumber=6163004
http://infoscience.epfl.ch/record/177415/files/Superpixel_PAMI2011-2.pdf
http://www.quyanyun.com/files/icpr/icpr/media/files/2592.pdf
http://ieeexplore-1ieee-1org-1ieee.rb-han.de.bosch.com/stamp/stamp.jsp?tp=&arnumber=6163004
http://www.6d-vision.de/9-literatur/pfeiffer_bmvc11
http://www.cvlibs.net/publications/Lenz2011IV.pdf
http://www-robotics.jpl.nasa.gov/publications/Arturo_Rankin/asc2006-rankin-water-detection-final.pdf
http://www-robotics.jpl.nasa.gov/publications/Arturo_Rankin/asc2008-rankin-mud-detection-final.pdf
http://ieeexplore-1ieee-1org-1ieee.rb-han.de.bosch.com/stamp/stamp.jsp?arnumber=06356185
http://klewel.com/conferences/epfl-computer-vision/index.php?talkID=7
http://lmb.informatik.uni-freiburg.de/Publications/2011/Bro11a/
http://cv.utcluj.ro/optical-flow.html
http://www.6d-vision.de/9-literatur/mueller_cvpr11
http://vision.middlebury.edu/flow/
http://www.cvlibs.net/datasets/kitti/eval_stereo_flow.php?benchmark=flow
http://www.6d-vision.de/9-literatur/stein_dagm04
Difficulty Comments
hard
medium
medium
hard
hard
hard
hard
medium
easy
hard
easy
Step by Step approach
Note: this has a generic formulation. The goal is to segment objects from stereo + image data by modifying standard
segmentation algorithms.
Can use http://www.cvlibs.net/datasets/kitti/eval_object.php as database/ground truth.
Good reference for segmentation methods in general: http://www.klewel.com/conferences/epfl-computer-vision/
0. K-means implementation
1. Standard superpixel method (ex: SLIC based on colour / graylevel and position) - use K-Means
2. Compute an initial disparity estimation (KTTI (http://www.cvlibs.net/datasets/kitti/eval_stereo_flow.php?
benchmark=stereo) or student implements a dense local method).
5. Joint segmentation (superpixels) and stereo estimation (see energy to be minimised in the paper for guideline) -
See MotionSLIC
Note: this is easier than the flow estimation using the same method - see next topic for flow
0. K-means implementation
1. Standard superpixel method (ex: SLIC based on colour / graylevel and position) - use K-Means
2. Compensate rotation using odometry.
3. Encode flow (epipolar flow) with a single parameter (VZ-index) and make the initial estimation via SGM.
4. Joint segmentation (superpixels) and flow estimation (see energy to be minimised in the paper for guideline).
0. K-Means implementation.
1. Standard superpixel method (ex: SLIC based on colour / graylevel and position) - use K-Means
2. Dense optical flow (ue existing public algorithms (http://www.cvlibs.net/datasets/kitti/eval_stereo_flow.php?
benchmark=flow) or student implements one)
3. Dense stereo (can be used an open source like the KTTI one).
4. Include in the metric info about the stereo and flow as in the reffered paper.
5. Group superpixels into objects (ex. use graph cuts, split and merge based on a superpixel descriptor etc.)