juliena82
commented
8 years ago As already discussed by mail, I would vote for option 3, with the nuance that we can keep the additionnal repo that Valera has already created instead of creating a new one, and also with the precision that all data necessary for the program (including downsampled version of images) would be in the main repository. For completeness, I include below my last mail:
- This is a small project. I have the philosophy to keep the things always as much simple as it can be (and I often fail at it, but I try). Having to manage 2 repositories is an added complexity that is really worth the separation of data and code? Think about a newcomer: he has to do 2 checkouts... Triste Think about us, we will have to do twice more as push as necessary in the future.
- Take for example OpenCV. This is not a small project. It is a big project. And even them have only one repository, containing all the source and images and videos for the sample (look at https://github.com/Itseez/opencv/tree/master/samples/data for instance). If the OpenCv developpers have done like this, I think that we should not be ashamed to do the same ;)
- You make a point when you speak of the potential problem of the big files. But I see it the other way: having only one repository will impose us to be reasonable on data size. For instance, when I included the HDR demonstration, I took at first the original images, but after I donwsized them and recoded them as JPG, so that they take less size.
So I let you decide, but if you take some time to think at this, I would be very glad...
My credo: always simpler (that's also why I think there is much clean up to do in the sources).
PS :
Still, I think that your idea to have a assets repository is a good idea and should be kept!
In fact, we could store here the original big data files that are candidate to be used for the demonstrator (before being downsampled / reduced / recoded in the ocvdemo/data final folder).
This way, we could have always access to the original data images or videos (even big).
What do you think of this?
valera-rozuvan
shervinemami
When I created the opencv-demonstrator-data [link] repository that contains the entire contents of the folder ocvdemo/data, I was thinking along the following lines:
However, I didn't take into the account that the ocvdemo/data folder also contained
.xmlfiles for the MMI (basically the GUI layout configuration). Those are required to build the project. So we need to bring back all of the.xmlfiles into opencv-demonstrator repository.What about the image assets? Keep them separate in another GitHub repository, and introduce a new folder called assets for them? Or sack the whole idea of a separate repository for the images (and later videos), and bring back the entire data folder how it was initially?
So we need to decide between 3 possible options.
Option 1
Bring back the ocvdemo/data only with the
.xmlfiles. The images will stay in a separate GitHub repository called opencv-demonstrator-assets, and will go to the folder ocvdemo/assets for the complete build. One good point about this approach brought up by @juliena82 in an e-mail discussion is this:Option 2
Bring back the ocvdemo/data folder completely. I.e. like it was in the beginning - all of the
.xmlfiles and images come along with the sources in the ocvdemo/data folder. Delete the opencv-demonstrator-data GitHub repository,Option 3
Like option Option 2, but we make another GitHub repository opencv-demonstrator-assets, and use it as a place to store high resolution images (and later videos). It can be used by package managers to create a separate package for the assets of this project.
So, @shervinemami and @juliena82 - let's vote on the best option to proceed forward! Please leave a comment to this issue with your choice = )