GSoC 2k18

[tosalonijain]

Considering that the application process for GSoC has already begun, I was wondering what are the top priority modules that are supposed to be wrapped during the course of GSoC timeline. Once known, I shall frame my proposal accordingly.

[tosalonijain]

@maxruby it would be better if I could get to know this as early as possible. Thankyou.

[maxruby]

Hi @tosalonijain

Thanks for your interest in OpenCV.jl. It would be nice to extend and further maintain this library. I would love to guide you but unfortunately, I do not have any time to invest on this right now or during the Summer.

Perhaps you could indicate which features of OpenCV interests you and I could suggest where to extend or improve the bindings for Julia.

[tosalonijain]

@maxruby, I wanted to ensure whether or not you are going to accept proposals for GSoC 2k18 in this project??

[tosalonijain]

@maxruby , As I can see OpenCV.jl presently has julia wrappers for the core, imgproc, videoio, highgui and video modules and is missing wrappers for gpu, ml, features2D, objdetect, calib3d, photo, flann majorly. Although it seems the already wrapped modules are not complete when compared to the list of features they offer in OpenCV. For example in OpenCV_core.jl Dynamic structures are not included, Basic structures and Operations are also not complete like SparseMat and few others also. One possible reason for this as I can sight is that whether or not Cxx is ready to handle the complexities associated with dynamic memory locations and pointers? I would like to work on core, highgui, imgproc ( further developing them) and include new modules preferably gpu, ml and rest that you may suggest. I am very much interested in taking this project and have started preparing my proposal for GSoC. It would be great if you could help a little. Thank you.

[maxruby]

Once you have your proposal ready and you share it with me, I will try to give you feedback as early as possible, but I as I said already, I will not be able to dedicate much time to this.

[tosalonijain]

Thank you @maxruby . I shall be very soon uploading the first draft of my proposal. Expecting suggestions from your side on that.

[tosalonijain]

@maxruby here is the link to my proposal. https://docs.google.com/document/d/1W4B-bRxPHKcorh5sQESPzIw2es1L2vtENWzvUielgU0/edit?usp=sharing Please review it and suggest required changes. I have uploaded on the GSoC site also. Thank you.

[tosalonijain]

@maxruby Please review once if possible. Thank you.

[maxruby]

@tosalonijain Sorry for the delay in getting back to you. I have superficially reviewed your proposal. Generally speaking, I find the proposal very good and it is nice to see your enthusiasm for working with OpenCV in Julia.

However, I think the proposal as it stands is overambitious (and not realistic), to a point where I believe that it needs to be trimmed down. I think updating OpenCV.jl to be compatible with the latest stable version of Julia (0.6) is in itself a significant goal from my recollection. In the past at least, we encountered various problems with arrays and array handling, and I think you could spend in principle a lot of time just to make sure that everything is compatible. The challenges you described regarding pointer handling in Cxx are also likely underestimated and could lead to delays in wrapping certain modules / functionalities.

Therefore, I suggest that you focus on maximum 5 primary tasks rather than trying to wrap a long list of missing features. For example:

• Improving documentation
• Update OpenCV.jl to be fully compatible with Julia latest stable release
• SparseMat support
• Image Queues
• GPU module integration (data structures, matrix operations)

[tosalonijain]

@maxruby Thank you for your genuine review. Yes I was myself a bit in dilemma whether or not the goal I have mentioned could be achieved in the time given. That's why I was desiring your review. Though, now as you have mentioned I would better focus on the above listed features. I was also suspecting if it would be a good choice to involve pointers with Cxx. Although I may take this as my post GSoC goal if it seems okay to you. On a frank ground, I wanted to ask what chances are there for this this project and I to be selected for GSoC 2k18?

[maxruby]

On a frank ground, I wanted to ask what chances are there for this this project and I to be selected for GSoC 2k18?

Honestly, I think the current project as it stands might perhaps not have enough merit in terms of its originality and scope to compete with others. However, this is purely my speculation and I would not want to discourage you from trying...

[tosalonijain]

@maxruby Thank you for the honest response. I just wanted to know, although I would greatly wish to be selected for this project and work on it.