2014 - alalek/opencv GitHub Wiki
Template is at https://github.com/opencv/opencv/wiki/Template
http://code.opencv.org/help/wiki_syntax
Agenda
- GSoC pre-pre-pre planning …
- Website
- Positions
- Intel
Minutes
- GSoC pre-planning
- We looked at who performed, who didn’t and why
- Some geographies will now require a programming test
- We will in general favor programming over vision. Given the short span of time, a student has to hit the ground running and it’s turned out that programming trumps vision knowledge. However, getting both is the ideal.
- We looked at who performed, who didn’t and why
- We are updating the website. It’s a tad dry. Vincent on this.
- Positions (send bios to Vincent)
- Gary — CEO
- Vincent — Treasurer
- Grace — Director of Development
- Vadim — Chief architecture
- Achin — Contact him. Role
- Grace has been dealing with Intel to get our payments straightened out
- We need to get more open, less SoW funding from them if possible.
- Contacts with other companies and DARPA in progress
Action Items
Gary
- Look into contacts with Caffe http://caffe.berkeleyvision.org
- Look into UC Berkeley grant
- Get bios
Vincent
- Website
Vadim
- .
Grace
- Intel
- Berkeley
- Bio
From Last Time
Gary
- .
Vincent
- .
Vadim
- .
Victor
- .
<><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><>
Agenda
- GSoC 2014, Feb 3 – 14 Org applications!
Minutes
- Need to figure out themes
- http://www.google-melange.com/gsoc/homepage/google/gsoc2014
- Code into OpenCV Contrib instead of main. Then 3rd party such as a linear algebra
- Code from Pieter Dollar segmentation
- Matlab
- http://vision.ucsd.edu/~pdollar/toolbox/doc/
- http://vision.ucsd.edu/~pdollar/files/papers/DollarICCV13edges.pdf
- http://research.microsoft.com/en-us/downloads/389109f6-b4e8-404c-84bf-239f7cbf4e3d/default.aspx
- Matlab
- Sparse Matices
- Segmentation (Dollar)
- Pedestrian detection, human detection
- Face detection — advance the state
- In opencv extra, large data set 24×24 (several 1000 faces)
- http://www.i-programmer.info/news/105-artificial-intelligence/5062-alien-facehugger-v-predator-face-tracking-hots-up.html
- http://www.micc.unifi.it/pernici/
- Soft cascade (half cascade in OpenCV)
- Needs to be improve
- Fastest pedestrian detection in the west
- training, new features, detection
- interframe interpolation and smoothing
- outlier supression
- Face detection — advance the state
- DTAM
- Tracking — predator (tracking, detection + learning)
- Was pretty rudimentary
- Draw box, resize
- Box distortion
- Look for candidates
- Was pretty rudimentary
- solvePnP update
- http://www-users.cs.umn.edu/~joel/index.php?page=software
- Optical flow
- State of art http://www.ingentaconnect.com/content/klu/visi/2014/00000106/00000002/00000644
- Prepare testing framework for quality
- Have frameworks for features and stereo already
- Improve Matlab wrappers from last year
- 3D —
- livmv finally integrated
+
- LSD
- Python
- computational photography
- Matlab
- Matas text detection (can detect road signs etc)
Action Items
Gary
- Set up the website for 2014. Use OpenCVGSoC13
- Contribute ideas
Vincent
- Contribute ideas
Vadim
- Contribute ideas
Grace
- Contribute ideas
From Last Time
Gary
- .
Vincent
- .
Vadim
- .
Grace
- .
<><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><>
Agenda
- GSoC we are applied, we find out Feb 24th
- Mentors
- Ideas https://github.com/opencv/opencv/wiki/GSoC_2014
- Toyota
- CVPR
Minutes
- Mentors
- Add alexander shishkov
- Add Stefano Fabri and his students
**
-
CVPR https://docs.google.com/document/d/1iEjpar0fO_BYwnQoSpXFJzqcT11Q1KhDblu8pV5G87g/edit
- http://www.pamitc.org/cvpr14/tutorials.php
- Start prep
- Who goes
- Tutorials ideas
- Tilt shift, full process to pull request
- ML: Linemod, softcascades, random forests/extremely
- OCR app, as a web app
- Need a coding test for students in GSoC. Simple one
Action Items
Gary
Transfer Stefano’s GSoC ideas across and his studentsWrite Ethan about
Vincent
- .
Vadim
- .
Grace
- .
From Last Time
Gary
- .
Vincent
- .
Vadim
- .
Grace
- .
<><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><>
Agenda
- GSoC 2014
- Foundations website changed
Minutes
- GSoC 2014
-
Our ideas page
- https://github.com/opencv/opencv/wiki/GSoC_2014
- Most projects should be done in OpenCV contrib. From there we can migrate top stuff to core
- In 3.0, we are trying to simplify core and put other things in contrib
- Sparse matrices — basic operations can’t do with them. Right now, they are just storage structures. They aren’t efficient for operations. Dense row or Dense column format for multiplication for example.
- Bundle adjustment would be good for the core
- Guide for making contribution
- https://github.com/opencv/opencv/wiki/How_to_contribute
- Docs, unit test, example, youtube videos.
- Good to just have projects generating example and more docs of existing code such as livmv
- Visualization: http://web.mit.edu/vondrick/ihog/
-
Our ideas page
- Mentors
- 2 more added to Wiki. 17 mentors with 2 backups
- Adding an object to describe a camera. But it’s hard because of
- Single camera
- Stereo
- RGBD
-
Foundation’s website changed
- http://vrabaud.github.io/foundation_site/
- Need ongoing funding for pull requests and admin functions
Action Items
Gary
- Write up Blog for GSoC “Good mentor/student story” Founded a company
- Funding: high school bring down. Intel. Google. Samsung. Buidu
Vincent
- .
Vadim
- .
Grace
- .
From Last Time
Gary
- .
Vincent
- .
Vadim
- .
Grace
- .
<><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><>
Agenda
- GSoC
Minutes
- Go through list https://groups.google.com/forum/#!forum/opencv-gsoc-2014
- Email to mentors to sign up
- Email Students to get applications in http://www.google-melange.com/gsoc/events/google/gsoc2014
- Really want
- http://www.robesafe.com/personal/pablo.alcantarilla/kaze.html
- 3D
- normals, features …
Action Items
Gary
- email to mentors
Vincent
- .
Vadim
- .
Grace
- .
From Last Time
Gary
Email to mentors “Please get on, choose projects …”
Vincent
Email to students “Please officially apply”
Vadim
- .
Grace
- .
<><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><>
Agenda
- GSoC
- Vadim back from mysterious “holiday” in the former Ukraine areas
- Grace still there on Vacation
- Payment received from …
Minutes
- GSoC. Some of note
- aKaze (features 2D)
- new PNP
- Python tutorial
- Computational photography → image decomposition
- Improve pictorial images. Felsenshwalb pictorial image, Deva Ramanam’s method
- P. Dollar
- Pedestrian detection
- Predator
- Anything relevant to person detection and tacking
- 3D stuff
- Fast normals, plane finder OpenCV extra
- 3D features or primitives
- Improve line mod in OpenCV
- Libmv
- Caffe
- Money is in from youknowwho. Can pay our OpenCV summer of code guy or his sister
Action Items
Gary
- Joining in deep learning, CNN. Contact them
Update ideas page
Vincent
- Final call for GSoC to Facebook
- Find out about paying our private OpenCV summer of code guy
Vadim
- Last reminder to the list
Grace
- .
From Last Time
Gary
Email to mentors “Please get on, choose projects …”
Vincent
Email to students “Please officially apply”
Vadim
First reminder to get in projects to mail list
<><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><>
Agenda
- GSoC
- … payment is in. Need to pay Itseez
Minutes
Only 2 items:
- Mentors have to sign up for OpenCV https://www.google-melange.com/gsoc/homepage/google/gsoc2014
- Interim period all mentors go to https://www.google-melange.com/gsoc/proposal/list/org/google/gsoc2014 and rate the proposals
Action Items
All
- Send out messages to mentors to sign up, and rank proposals
Gary
- .
Vincent
- .
Vadim
- .
Grace
- .
From Last Time
Gary
- .
Vincent
- .
Vadim
- .
Grace
- .
<><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><>
Agenda
- GSoC
- OpenCV mini-challenges
Minutes
GSOC
- Priority to students who performed well in previous years
- Soft cascades patented, need to use close variant
- Boosting chain, better
- aKaze
- Scene text detection Louise Gormez
- We will have a special mentor this meeting to do student selection.
OpenCV Mini-Challenges
- Funding to support interaction
- Choose features/goal/classifiers to work on
- Prize money
Action Items
Gary
- .
Vincent
- .
Vadim
- .
Grace
- .
From Last Time
Gary
- .
Vincent
- .
Vadim
- .
Grace
- .
<><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><>
OpenCV GSoC 2014 students, projects, mentors:
Student Project Mentor(s) Lluis Gomez Improve Scene Text Detection module in OpenCV Stefano Fabri, Vadim Pisarevsky Abid Rahman Extend of OpenCV-Python Bindings and Tutorials Vadim Pisarevsky Fedor Morozov Local feature descriptors Pablo Alcantarilla Siddharth Kherada Computational Photography: Image Decomp Color spmallick Daniel Angelov Custom Calibration Pattern Sid Bao Vitaly Lyudvichenko New edge-aware filters for OpenCV Alexander Mordvintsev Alex Leontiev Learning based trackers Gary Bradski Tolga Bridle Rec & Pose Est of 3D Objects via 3D Features Bence Magyar Francesco Puja SALIENCY IMPROVEMENTS FOR TRACKING Antonella Cascitelli edgarriba Real Time Pose Est Tutorial and DLS impl. Alexander Shishkov Yuri Gitman P. Dollar segmentation algorithm Alexander Shishkov Vlad Shakhuro Fastest Pedestrian Detector in the West Alexander Bovyrin biagio Geometrical-primitives based localization system Manuele Tamburrano Tomasz Pietruszka Optical Flow Estimation – state-of-the-art Prasanna Krishnasamy Paul Foster Dense Tracking and Mapping (DTAM) for OpenCV Vincent Rabaud
- Bill DARPA
- Help set up contests
<><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><>
Agenda
- GSoC
- CVPR
- OpenCV 3.0
- Kickstarter
Minutes
- GSoC
- Some students already moving
-
CVPR
- need to register
- Vadim can’t attend
- Will deliver slides
- Need to look for extra support
- OpenCV 3.0
- Beta by CVPR
- Sept final thing
- Kickstarter
- Simple video
- Shave our heads until we make our goal
Action Items
- Mentor email
- all in touch. all confirm
- Set up for CVPR workshop
From Last Time
<><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><>
Agenda
- Who is going to the mentor summit?
- My email https://mail.google.com/mail/u/1/#inbox/145f210b78e5342f
-
CVPR workshop
- Wrote Cat
- Go over challenge file
- https://drive.google.com/file/d/0B4UgujoQOxIAX25ZbUgtTUhZSVgxSTlKV3JSUlBCTDBKNnVn/edit?usp=sharing
Minutes
- Mentor summit
- A Shishkov
- Gary
-
CVPR
- Contests
- Most want to implement conference vote
- Put up workshop page on wiki
- Get schwag
- Fill in high level outline slides of workshop.
- Coding starts May 19
- We received PO from Google
Action Items
Gary
Mentor and Student coding startSet up common report docsVideos set upStart wiki page for coding
Vincent
- .
Vadim
- .
Grace
- .
<><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><>
Agenda
After a long hiatus for summer vacations, we’re back
- Functional Programming
- Machine learning
- OpenCV Contest
- GSoC 2014
Minutes
- Functional Programming
- We will look into working with Kunle Olukotun’s Pervasive Parallel Computing group to begin exploring a functional coding of OpenCV/Vision to enable long term optimization and keeping up with new types of hardware
- Vadim is developing a functional language taken from others that specializes in array comprehension
- Machine learning
- The machine learning side of OpenCV is a bit dated/stale
- Vadim has recoded much of it (not some of the boosting) and this will begin to show up in 3.0
- We are also going to work on taking learned deep learning architectures expressed as XML files and compiling them to OpenCL for optimal runtime performance.
- OpenCV Contest
- We now have a full time programer, Dimitry to support the OpenCV contest.
- He is working on a standard way to access image datasets
- We will promote this way/format to others so that datasets may have more uniform access in the future
- Basically code or translate via script the data to an XML format and we’ll provide standard reader/writers from there
- GSoC 2014
- We need to collect all the student videos
- Gary needs to write the blog
- Finalize on the mentor summit
Action Items
Gary
Finalize mentor summit- Write GSoC blog
Contact KunleContact SilvioContact Fei-FeiTransfer my BMVC poster etc to Grace
Vincent
- Edit videos
Vadim
Send links to videos he knows about
Grace
- Assemble videos
- Present at BMVC
From Last Time
Gary
- .
Vincent
- .
Vadim
- .
Grace
- .
<><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><>
Agenda
- GSoC analysis
- Functional coding
Minutes
- OpenCV GSoC 2014
- Most successful summer ever: 15 students resulting in 14 pull requests
- A pull request this year includes documentation and videos
- Gary, Vadim will write up what was accomplished for the Google Summer of Code blog
- Vincent will edit the contributed videos into a retrospective
- Summit attendees
- Gary
- Alexander
- Most successful summer ever: 15 students resulting in 14 pull requests
- Functional coding
- This effort is to improve the code base of OpenCV so that:
- There will be one code source only
- The code, once compiled, tends to be much more robust against bugs, corner cases
- Functional programs are much easier for compilers to optimize
- To C++, to Python
- To OpenCL
- And to embedded architectures which are proliferating and hard to keep up with. This way, the manufacturer can make sure the compiler will optimize code to their own architecture.
- Vadim will start work on this October 15
- This effort is to improve the code base of OpenCV so that:
Action Items
Gary
- Write GSoC 2014 blog entry
Vincent
- Edit the videos from the students
Vadim
- Help bring out the utility of each GSoC entry
Grace
- Get videos to Vadim
Invoice to Berkeley for work on coding contest
<><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><>
Agenda
- OpenCV Challenge
- GSoC Video
- OpenCV 3.0 for ROS
Minutes
- OpenCV Challenge
- Data set readers prepared
- Todos
- Establish approximate state of the art thesholds
- Update https://docs.google.com/a/magicleap.com/document/d/1aCKmUpwpDxozC7cmIcokKRqQdcA9cckXTb1G8WOLYmI/edit#
- By Friday, get the document ready
- Announce the contest by Monday
- Contestants must meet or exceed, self reported
- Among the top 5 (by our jury decision), we will verify their code and announce the winners
- Those who submit accepted OpenCV pull requests, will get large allocated amount, otherwise they’ll get $100
- Vincent is working on the GSoC 2014 Video (14/15 resulted in accepted pull requests), but it will not be ready this week
- ROS is being updated to use OpenCV 3.0!
Action Items
Gary
- Figure out state of the art
Vincent
- .
Vadim
- .
Grace
- .
From Last Time
Gary
- .
Vincent
- .
Vadim
- .
Grace
- .
<><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><>
Agenda
- OpenCV Challenge
- GSoC Video
Minutes
- 2.4.10 : out this week
- opencv tested on ARM. Some NEON optimizations
- 2.x: support for Java (better: smart pointers are handled)
Action Items
Gary
- .
Vincent
- DARPA announcement
- Contact Qualcomm
Vadim
- .
Grace
- .
From Last Time
Gary
- .
Vincent
- .
Vadim
- .
Grace
- .
<><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><>
Agenda
- GSoC
- Contest: Prepare state of the art.
- 5 categories for challenge
- Select such for minimum support needs
- SLAM 1.3%
Minutes
- GSoC
- Alexander coming to the mentor summit
- Need to finalize invoice to Google
- Contest
- Focus on gestures, 3D data
- Gary to send list
- Dimitry to come up with graphics simulated examples
- https://docs.google.com/a/magicleap.com/document/d/1aCKmUpwpDxozC7cmIcokKRqQdcA9cckXTb1G8WOLYmI/edit#
Proposed gestures list:
- Thumbs up, down, right, left
- “L” facing away
- “L” facing towards
- Fingers up facing away: 1, 2, 3, 4 (separated)
- Fingers up facing towards: 1, 2, 3, 4 (separated)
- Open palm away (fingers separated)
- Open palm towards (fingers separated)
- Open palm away (fingers lined up)
- Open palm towards (fingers lined up)
- Open palm held out (fingers separated)
- Open palm held out (fingers lined up)
- Pinch
- Fist closed facing away
- Fist closed facing towards
- QC
- Snapdragon optimization
- OpenCV 3.0 => FastCV
- Binary
- Run time switch
**
Action Items
Gary
- Send list of gestures
Vincent
- GSoC Videos!
Dimitry
- Graphically generated examples
Grace
- .
<><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><>
Agenda
- Contest
Minutes
- Contest
- Now have 13 datasets that can be read
- Contact
- 3.0
- To do
- Primary goal is to finalize the API, some not backward compatible
- Basic functionality API will stay unchanged from 2.X, load, store …
- High level interfaces will change sightly
- Reference manual switching to Doxygen
- Place general ref in dox that … similar to RST
- Alpha is released
- Beta end of October
- Full release: End of year
- To do
Action Items
Gary
- .
Vincent
- .
Vadim
- .
Grace
- .
From Last Time
Gary
- .
Vincent
- .
Vadim
- .
Grace
- .
<><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><>
Agenda
- Contest
- Code
- Our site is down
Minutes
- Site is down. Shishkov is back on Thursday to repair
- Contest info https://docs.google.com/document/d/1B3q2uDDwKlfen8nIPGUYl_j3cXK3mjar1USUrEUpGLA/edit#heading=h.iakcszekqaak
- Contest information is at: https://docs.google.com/document/d/1B3q2uDDwKlfen8nIPGUYl_j3cXK3mjar1USUrEUpGLA/edit#heading=h.iakcszekqaak
- Have access to all 21 datasets (19 datasets plus one replacement plus on tracking set that has its own validation framework)
- Documentation and examples done
- Need a few more days to finish scripts
- Documentation and examples done
- Need to wait to Thursday for site to come back up
- Try to announce contest on Friday
- On Thursday, send out emails to verify
- OpenCV 3.0
- Close to release of 3.0 Beta
- Release first week of Nov
- National unity day holidays in Russia
-
ARM optimization, bug fix, pull requests,
- Most new functionality in OpenCV contrib
- Need notes on how to do this add contrib on build
- See change log
- Need notes on how to do this add contrib on build
- Most new functionality in OpenCV contrib
- Documentation is next concentration, switch to doxygen
- Maxim is adding doxygen for 3.0 final
- Close to release of 3.0 Beta
- Functional recode
- https://docs.google.com/spreadsheets/d/1dL2MYJpJ6AVnggtsrX0XjzF9cjajUgoy6Ta1jfqLdaU/edit#gid=0
- Goal to produce programming efficiency WITH optimal code
- Was out sick with flu
- Spent time on the language
- Major StandardML
- Close to C++ syntax
- Major StandardML
- Stages
- 1st Language speed like java, mainly functionality, better than python. Compiler for subset such as object training
- 2nd templates loop and array comprehension … might need some cross platform intrinsics (on Intel, ARM …)
- 3rd optimization loop fusion for speedup, expand inline functions — machine independent .
- 4th optimization is to hardware — go to “enriched C with intrinsics” (Neon, SSE …)
- Stage 1, EOY
- Git Magic Leap
- Apache 2.0
- Target for this
- At least as powerful as Matlab
- Achieve high efficiency code, C level, especially on mobile
- Target other platforms: GPU
*
Action Items
Gary
- Send out emails on Thursday for contest
Vincent
- .
Vadim
- .
Grace
- .
From Last Time
Gary
- .
Vincent
- .
Vadim
- .
Grace
- .
<><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><>
Agenda
- PO
- Contest
- OpenCV.org back up, was spammed.
Minutes
- Contest
- Need to be announced this week
- Announcement proposal https://docs.google.com/document/d/1aCKmUpwpDxozC7cmIcokKRqQdcA9cckXTb1G8WOLYmI/edit#heading=h.mpvtirb42db1
- Datasets https://docs.google.com/document/d/1B3q2uDDwKlfen8nIPGUYl_j3cXK3mjar1USUrEUpGLA/edit#heading=h.iakcszekqaak
- Need to be announced this week
- Since numl is much closer to StandardML (SML) than to OCaml, and I’m not proficient in either, and since we just started, I decided to spend a few extra days and convert the whole existing compiler code from OCaml to SML. The work has been nearly finished and now I’m debugging the extended parser (that supports new features, such as obligatory type declarations and such):
- OpenCV 3.0
- Nov 7 release of 3.0 Beta
- new optimization
- OpenCL optimization
- GSoC 2014 results included in this release (in contrib and in the core part)
- AKAZE, Computational Photography bits
- Nov 7 release of 3.0 Beta
- OpenCV 4.x work is started in functional language. Similar to StandardML (SML). First version due in December
Action Items
Gary
- PO to Google for GSoC and Mentor summit
Vincent
- .
Vadim
- .
Grace
- .
From Last Time
Gary
- .
Vincent
- .
Vadim
- .
Grace
- .
<><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><>
Agenda
- OpenCV 3.0 Beta
- GSoC
- Contest
Minutes
- OpenCV 3.0 Beta
- Release is out: http://opencv.org/opencv-3-0-beta.html
- Final release early Jan
- 40 new functions optimized on Neon (resize, color converstion, arithmetic …)
- Before EOY, release final candidate will come out
- 14 GSoC projects included
- GSoC
- Need to invoice Google
- Have travel expenses, need to reimburse
- Contest
- Need to get the announcement out
- Have Dmitry to support
- Hope to send this out on Nov 13, 2014
- Functional language
- Preparing Beta
- Parser
- Plans
- OpenCV benchmarks in progress
**
- OpenCV benchmarks in progress
Report from Vadim
Numl: I continue stabilizing and extending the new version of Numl compiler, rewritten in StandardML. The lexer and the parser part have been debugged, they seem to handle small test programs correctly. The next step will be to add the module support and other new language features planned for phase 1. https://github.com/vpisarev/numl/tree/sml_test_parser OpenCV (put here for completeness, even though it's not the part of my paid contract): beta 3.0 is out: http://opencv.org/opencv-3-0-beta.html
Action Items
Gary
- PO to Google for GSoC and Mentor summit
- Contest announcement needs to go out
Dmitry
- Se
Vadim
- .
Grace
- .
From Last Time
Gary
- .
Vincent
- .
Vadim
- .
Grace
- .
<><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><>
Agenda
- Contest
- OpenCV 3.0 is out in Beta!
Minutes
- OpenCV 3.0 Beta http://opencv.org/opencv-3-0-beta.html
- OpenCV 3.0 Final should be out in January
- Contest
- Get contest out as widely as possible
OpenCV Vision Challenge OpenCV Foundation with support from DARPA and Intel Corporation are launching a community-wide challenge to update and extend the OpenCV library with state-of-art algorithms. An award pool of $50,000 is provided to reward submitters of the best performing algorithms in the following 11 CV application areas: (1) image segmentation, (2) image registration, (3) human pose estimation, (4) SLAM, (5) multi-view stereo matching, (6) object recognition, (7) face recognition, (8) gesture recognition, (9) action recognition, (10) text recognition, (11) tracking. Conditions: The OpenCV Vision Challenge Committee will judge up to five best entries. You may submit a new algorithm developed by yourself or your implementation of an existing algorithm even if you are not the author of the algorithm. You may enter any number of categories. If your entry wins the contest you will be awarded $1K. To win an additional $7.5 to $9K, you must contribute the source code as an OpenCV pull request under a BSD license You acknowledge that your contributed code may be included, with your copyright, in OpenCV You may explicitly enter code for any work you have submitted to CVPR 2015 or its workshops. We will not unveil it until after CVPR. Winners and prizes are at the sole discretion of the committee. Timeline: Submission Period: Now - May 8th 2015 Winners Announcement: June 8th 2015 at CVPR 2015 Detailed information can be found here: https://github.com/opencv/opencv/wiki/VisionChallenge Questions and the results should be sent to: [email protected] Thank you, OpenCV Vision Challenge Committee
Numl: No noticeable progress during last week; I concentrated on the plans on OpenCV 3.0. Will resume the work on the language (and hopefully spend full 2 weeks on it) starting this Wednesday. OpenCV: 3.5 documents have been prepared on the OpenCV-related tasks (the document on HAL and optimizations is in progress)
Action Items
Gary
- Write Hinton about X-Prize
- Find out what Baidu wants/could do.
Vincent
- .
Vadim
- .
Grace
- Get in GSoC invoice
From Last Time
Gary
- .
Vincent
- .
Vadim
- .
Grace
- .
<><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><>
Agenda
- Contest
Minutes
- Contest: announcement is out
- Spread in Europe and China/Asia
- Contacts for OpenCV
- Anan Peleg
- Visual computing institute: Europe
-
NIST
- OpenVX is a possibility
- 40 algorithms
- OpenVX is a possibility
- OpenCV 3.0 Beta was released last week
- Plans for 3.0 final, end of December
- January for 3.0 final release
- Reference manual converted to Doxygen
- http://docs.opencv.org/ref/master/
- Plans for 3.0 final, end of December
- Qcom fastcv discussion
- NVidia Representitive for Europe Marketing
Vision Challenge We (OpenCV Foundation) are running a computer vision contest. The goal is to get state of the art algorithms in at least 5 of 11 vision categories. We will have a workshop at CVPR 2015 to go over the winning entries. The code will then be available in OpenCV contrib. It is OK to duplicate existing state of the art in an OpenCV compliant framework under a BSD license or to develop new algorithms/code. It is OK if you have submitted the work elsewhere or currently submitted to this CVPR, we will not leak entries early. This contest is for anybody, or any group, corporate, academic or private. There will be a $50K award pool to distribute to the winners in the 5 top categories. Please distribute and encourage people you know to enter! You do not have to attend CVPR to enter the contest. http://opencv.org/opencv-vision-challenge.html Thank you Gary Bradski CEO OpenCV
Progress report
Numl:
After the delay related to OpenCV 3.0 beta release and the further plan development, the work on numl was resumed.
Learnt the useful MoscowML module PP (pretty printer generators) and implemented the pretty printer for the Numl
abstract syntax, which should be very useful to debug various compiler steps. The module can also be used to generate
the final nicely indented C code out of its hierarchical representation.
Using the AST pretty printer checked that the lexer and parser work fine after conversion to MoscowML.
Then, at the lexer/parser level added the following new stuff (this one is still not finished):
many primitive operations (almost complete set of Matlab/C++ operations, except for augmenting “+=” etc.)
more primitive types: single- and double- precision floating-point, 8-, 16-, 32-bit signed and unsigned integers, text strings
complex types: references, lists, multi-dimensional arrays, records, tagged variants and the syntax to define new types
mutually recursive functions
basic syntax for defining and throwing exceptions (no exception handling yet)
extended “if” statement with zero or more "elif"’s and optional “else”.
import statement for using other modules
The latest snapshot is available here: https://github.com/vpisarev/numl/tree/sml_test_parser/srcNuml Plans:
add some more elements to the lexer and parser:
switch (pattern matching) statement
exception handling syntax
accessing array slices
array literals
external c functions
finish such a parser, check it using updated pretty printer
parse the imported modules
implement syntax verification and type checking part, at least for the subset of the syntax
implement conversion to the simplified (k-normal) AST representation
after all this is done, the only 2 non-trivial things to implement would be the C code generator and the runtime.
all the other components can be adopted from the min-caml sources.OpenCV:
Finished the document on possible OpenCV optimizations
Maxim Shabunin continues the work on semi-automatic RST=>Doxygen conversion. Some first results are available
here: http://docs.opencv.org/ref/master/
Action Items
Gary
- .
Vincent
- .
Vadim
- .
Grace
- .
From Last Time
Gary
- .
Vincent
- .
Vadim
- .
Grace
- .
<><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><>
Agenda
Minutes
- Contest:
- Dmitriy: more contacts with people that participate
- answer more questions to help people
Action Items
Gary
- .
Vincent
- .
Vadim
- .
Grace
- .
From Last Time
Gary
- .
Vincent
- .
Vadim
- .
Grace
- .
<><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><>