DCV revived

December 13, 2021

DCV revived

Posted by Ferhat Kurtulmuş

Permalink

Ferhat Kurtulmuş

Permalink

After some effort, I could make DCV compilable with the recent versions of LDC, mir libraries, ffmpeg, bindbc.opengl, and bindbc.glfw. It still has some problems:

stereo vision module needs a revision I could not figure out so far.
ggplot backend must be re-implemented or something better can be used. This one is just for visualization though. It would be awesome to be able to use Adam's simplewindow instead of glfw, but it requires so much work.
Reading from video files causes distorted images for some video types. Some test video files work very well. There must be a stride problem or unimplemented pixel format in https://github.com/aferust/dcvrev/blob/main/video/dcv/io/video/common.d#L225 (adoptFormat). Reading from camera devices seems fine so far.

I also contributed to it with a new module "measure" covering some common binary region algorithms such as label, contour, ellipse fitting, convex hull, and raw moments. Some of them are translated from the scikit-image Python library mostly written in cython.

I used ref-counted Slice!(RCI:T, N, kind) in the newly-implemented functions instead of GC-allocated slices. I don't know if it is a good idea though. These two types must work together somehow. For now, they are doing well IMHO.

I am not familiar with the implementations of every single algorithm previously-implemented in DCV. So, I am not sure about undertaking the whole thing. I think I will keep it as a separate repo on me for a while, and let's see what will happen.

What does work: every example in the example folder except the unrevised folder that includes examples requiring ggplot and multiview (stereovision) module.

Please contribute. I can even transfer the entire repo on request. What I want is to keep it alive.

https://github.com/aferust/dcvrev

On 13.12.2021 15:48, Ferhat Kurtulmuş wrote: > After some effort, I could make DCV compilable with the recent versions of LDC, mir libraries, ffmpeg, bindbc.opengl, and bindbc.glfw. It still has some problems: > > * stereo vision module needs a revision I could not figure out so far. > > * ggplot backend must be re-implemented or something better can be used. This one is just for visualization though. It would be awesome to be able to use Adam's simplewindow instead of glfw, but it requires so much work. > > * Reading from video files causes distorted images for some video types. Some test video files work very well. There must be a stride problem or unimplemented pixel format in https://github.com/aferust/dcvrev/blob/main/video/dcv/io/video/common.d#L225 (adoptFormat). Reading from camera devices seems fine so far. > > I also contributed to it with a new module "measure" covering some common binary region algorithms such as label, contour, ellipse fitting, convex hull, and raw moments. Some of them are translated from the scikit-image Python library mostly written in cython. > > I used ref-counted Slice!(RCI:T, N, kind) in the newly-implemented functions instead of GC-allocated slices. I don't know if it is a good idea though. These two types must work together somehow. For now, they are doing well IMHO. > > I am not familiar with the implementations of every single algorithm previously-implemented in DCV. So, I am not sure about undertaking the whole thing. I think I will keep it as a separate repo on me for a while, and let's see what will happen. > > What does work: every example in the example folder except the unrevised folder that includes examples requiring ggplot and multiview (stereovision) module. > > Please contribute. I can even transfer the entire repo on request. What I want is to keep it alive. > > https://github.com/aferust/dcvrev Thank you for your efforts!!!

Forums