OpenCV, Computer vision for Google Summer of Code (GSoC) 2011
Object recognition with OpenCV, to be published in ICRA 2011
Next important dates:
- April 8: You are now too late to apply, see you next year.
. . .
- April 8: You are now too late to apply, see you next year.
- April 22:
- All mentors must be signed up by midnight (PDT) April 22.
- All students decided on by 10am (PDT)
- IRC meeting, freenode (use XChat) #gsoc 12Noon (PDT)
Times:
- UTC to PDT (Pacific Daylight Time == Daylight savings time. In winter it's PST).
For students interested in applying
To apply to us:
Go to gsoc2011
- Click on "My Dashboard"
- Click on "apply"
- After the organizations load (takes some time), type in "OpenCV", click on what comes up and apply.
Project Idea summary (see details under each mentor)
Students may propose their own projects, but below are some of our priorities (not in any order) for this year
OpenCV GPU Acceleration using OpenCL
A new module, similar to opencv-gpu, is to be created. It will include a Mat-like matrix type + a few OpenCL-acclerated functions, such as: stereo correspondence, face & object detection, FAST detector, computing and matching BRIEF/r-BRIEF descriptors. The student must have a good knowledge of C/C++, some acquaintance with computer vision, parallel programming & SIMD, and have access to an OpenCL-capable working environment.
Look at existing ports of OpenCV to iOS, decide if they are complete, find the "best" one.
- Contact authors of the best port and see how they plan to maintain it.
- Help pull the existing port into the OpenCV build/test system or else, write a better port of OpenCV to iOS
- Write example iOS apps for things like AR, calibration, panoramas etc.
- Document how to write these apps and how the apps may be used as code stubs to build on
- Make efficient converters and unify data structures if possible between the Point Cloud Library (PCL) and OpenCV
OpenCV and Structure from Motion (SfM)
- Implement as many Structure from Motion algorithms in OpenCV
- Use new, joint features approach to visual object tracking.
- Create an image collage program using many of the tools from OpenCV to automatically find images that go together, blend them, segment out objects, find faces etc etc.
- Document it and design it so that its easy for others to extend the functionality
- Possibly allow this to run on iOS and/or Android
- Use the 2D image and 3D point cloud from Kinect together with a 3D model of a person to do real time body detection and tracking
Take the many samples in .../opencv/samples/ and
- Document them so that people can see what is there, what each does and how to run them. Much of this info will print out if you run each example program without any command line parameters.
- Organize the sample code into coherent tutorials: Getting started, finding features, camera calibration, stereo, tracking, machine learning ...
- Fill in/add new samples
OpenCV image labeling with Amazon's Mechanical Turk
- We'd like to have an easy way for computer vision people to use Amazon.com's Mechanical Turk system for image labeling
- Take some learning algorithms such as BiGG, VFH, TOD, FLANN and make it easy to scale object detection on the cloud.
Stretch goals:
- Can you make an Android or iPhone app to do vision labeling apps with Mech Turk?
- Make it easy to start using OpenCV vision algorithms on Amazon's servers "S3"
Easy to hook up cell phone vision with, say, vision recognition apps on Amazon's S3 servers
- Take the existing C/C++ samples and rewrite them in Python,
- create some new ones such as image stitching, simple gesture recognition etc. The student must have a fairly good knowledge of Python and computer vision background.
Deep Learning/Sparse Autoencoders
Implement the Sparse Autoencoders described http://www.stanford.edu/class/cs294a/handouts.html
- Try out on some image recognition project TBD
Mentors
In the below, get rid of the -delete- to make the emails work.
Alexander Bovyrin PhD, Senior Researcher Argus/Itseez founder NNU Lecturer -delete-alexander.bovyrin@-delete-itseez.com
Alexander's Suggested Project:
Gary Bradski Senior Scientist, Willow Garage Consulting Prof. Stanford U. OpenCV Founder, Technical Content Owner, GSoC Admin http://www.willowgarage.com/pages/people/gary-bradski-senior-researcher -delete-garybradski@-delete-gmail.com
Gary's Suggested Projects:
Victor Eruhimov OpenCV founding team/Senior Researcher Argus/Itseez founder NNU Lecturer -delete-relrotciv@-delete-googlemail.com
Victor's Suggested Project:
Caroline Pantofaru Researcher Scientist in perceiving people http://www.willowgarage.com/pages/people/caroline-pantofaru-research-scientist -delete-pantofaru@-delete-willowgarage.com
Caroline's Suggested Project:
Vadim Pisarevsky OpenCV founding team/Czar -delete-Vadim.-delete-Pisarevsky@-delete-gmail.com
Vadim's Suggested Projects:
Vincent Rabaud Research Engineer/Perception, Willow Garage -delete-vrabaud@-delete-willowgarage.com
Vincent's Suggested project:
Ethan Rublee Software Staff/Perception, Willow Garage Ported OpenCV to Android See Android vision apps under "Robot view" or "the vegan robot" -delete-erublee@-delete-willowgarage.com
Ethan's Project:
Back up Mentors
Mark Asbach Fraunhofer IAIS Schloss Birlinghoven Sankt Augustin, Germany http://mmprec.iais.fraunhofer.de/asbach.html -delete-mark.asbach@-delete-iais.fraunhofer.de
Nicolas Saunier, Ph.D. Assistant Professor Civil, Geological and Mining Department (CGM) École Polytechnique de Montréal http://nicolas.saunier.confins.net -delete-nicolas.saunier@-delete-polymtl.ca
OpenCV Org. Application to GSoC 2011
Previous Years
Last year's OpenCV GSoC page is here