Ask Your Question
2

Move * opencv module in opencv_contrib

asked 2018-11-07 06:15:04 -0600

LBerger gravatar image

updated 2018-11-09 08:55:17 -0600

There is this question http://answers.opencv.org/question/20... and now (https://github.com/opencv/opencv/pull...)

( 1. Cuda move to opencv_contrib )

  1. Shape move to opencv_contrib

  2. videostab move to opencv_contrib

  3. superres move to opencv_contrib

  4. matlab script move to opencv_contrib

  5. move viz to opencv_contrib

  6. moved DIS optical flow from opencv_contrib to opencv, moved TVL1 in the opposite direction

What is the goal ? Is it finished ?

@mshabunin

edit retag flag offensive close merge delete

Comments

4 . 5 . -- long overdue ! (imho)

berak gravatar imageberak ( 2018-11-07 08:37:01 -0600 )edit

2 answers

Sort by ยป oldest newest most voted
3

answered 2018-11-16 15:01:02 -0600

mshabunin gravatar image

The thing is core team can not support so many modules in main repository, so we decided to move less popular (or useful) modules to opencv_contrib. We also can not support CUDA on a regular basis, because we don't have enough resources and expertise in this constantly evolving area.

Also, we wanted to move some algorithms to the main repository so that they become a mainstream thing, a basis; DIS optical flow is a nice example of effective modern algorithm with good quality.

Another fact is that in some areas traditional CV algorithms can not compete with DL-based approach, nowadays object detection can be performed using dnn module with better effectiveness than cascades.

cmake scripts for matlab has been moved to the corresponding module because it is the most appropriate place for them, my personal opinion is that we should spread all 3rdparty dependencies to corresponding modules, e.g. parallel backends to core, image codecs to imgcodecs, etc.. Currently build scripts have some legacy stuff from 2.4.x and many constructions to support older cmake versions (2.8.12).

We want to keep source compatibility, so moving from main repository is finished, but reverse still can happen.

edit flag offensive delete link more

Comments

I am interested in CUDA. Can I somehow help the core team ?

xxxcucus gravatar imagexxxcucus ( 2018-12-03 01:51:46 -0600 )edit

Yes, you can test and review existing PRs, verify and fix (build & test) CUDA algorithms on different platforms, with different toolkits and hardware. We recommend making PRs to the 3.4 branch of the main repository - we will port them to the master branch of opencv_contrib.

mshabunin gravatar imagemshabunin ( 2018-12-03 05:11:18 -0600 )edit
1

answered 2018-11-09 08:52:19 -0600

LBerger gravatar image

I find an answer to my own question here

Opencv 4.0

    DNN integrated into core as a central function
    Somethings move to conrib

๐Ÿ˜…

edit flag offensive delete link more

Question Tools

2 followers

Stats

Asked: 2018-11-07 06:15:04 -0600

Seen: 460 times

Last updated: Nov 16 '18