I am new to opencv and after an extensive search I have not found any detailed explanation to how one can implement a homograph. For example if I wanted to say... have a side shot of a chess board or a book and wanted to make a new image look as if the original were shot head on. I would try providing some code to show I attempted a solution but outside of knowing some of the functions I would need I am not sure how to implement them to solve my problem.
I am using python 2.7
thank you,
Related
So I have been assigned a project on quite short notice, and given that I have a very small amount of experience with c++ I thought it would be a good idea to ask for some help from the community!
Basically, we have a .png which is an image of a baseball game, and we must use opencv (and its likes) to detect, outline and extract coordinates of the bat, using c++.
So far I have installed opencv onto visual studio 2019 and read through hours and hours of tutorials but feel like I have not made a huge amount of practical progress.
As a beginner, I'm wondering if there is anyone here who could make the path forward a bit more clear for me!
Template machine is one of the easy solutions, below is an example for template matching from open cv -
https://github.com/opencv/opencv/blob/master/samples/cpp/matchmethod_orb_akaze_brisk.cpp
You can do this with many other python library's -
https://scikit-image.org/docs/dev/auto_examples/features_detection/plot_template.html#sphx-glr-download-auto-examples-features-detection-plot-template-py
This will not solve object detection problems in general but will be able to find an object of your interest in many cases, this is one of the quick solutions that I can think of.
I'm about to program a project in which I need to extract the straight lines out of an image.
The first thing that came to my mind is Hough Transformation. So I downloaded the OpenCV framework, added it to my project. Added the Bridging Header as well as the OpenCVWrapper.h and the OpenCVWrapper.mm files. But I can't find any tutorial on how to implement the hough transformation into my swift project. So what sort of files do I need to create to write the function in? And how should that function look like?
I'm happy to hear any helpful suggestion!
Kind regards
Robert
I'm also currently working on OpenCV with Swift, it indeed didn't have much tutorial about it.
About the Hound Transformation, here is the link to the documentation, hope it helps.
And this project helps a lot when I want some reference.
Good luck!
I'm very much trying to do what was asked here:
http://www.qtcentre.org/threads/45028-Designing-a-Node-Editor-(Hint-Blender-Node-Editor)
Which is basically a blender style node editor using Qt. Creating rects with circles which can be drag/dropped is fairly simple to figure out (see the image below).
The issue I have is how the algorithm for connecting the circles between nodes would work. Also I don't know how this could be drawn efficiently using QGraphicsItem API's? I have seen QPainter::cubicTo() which looks like it may be the way to go?
I assume it somehow manages to find a none overlapping path? But it must also handle the case where it has to overlap?
And what about the user being able to move these lines around in case the algorithm has done a bad job? I don't think this would be simple to implement using QPainter::cubicTo() as you could only move the curves control points?
I don't think you'll get anyone posting the code for doing the whole lot here.
Well, I volunteer -- just in case someone still needs such a library.
It is a Qt-based node editor implemented on top of QGraphicsView.
A library consumer defines data models, the rest is done by the node editor.
Some features:
Model-view approach.
Models describe data types, number of input and outputs and optional embedded QWidget.
The Scene could be saved to file.
The rest similar projects were either not finished, not supported anymore or lacking some features.
https://github.com/paceholder/nodeeditor
I would suggest that you implement it without taking into consideration overlap but with the possibility of overriding how the connecting line is drawn.
It looks like there were some good replies on the Qt forum. I don't think you'll get anyone posting the code for doing the whole lot here.
Perhaps take a look at the Qt demos Graphics View/Elastic Nodes - There is code for the demo that you can use as a starting point, but it is a lot of work to get from there to something like the Blender node editor.
Another node editor to take a look at is SynthEdit or Synth Maker (has a nice one).
I am to make a little demo app to learn some Qt and c++ and if anyone know some guides that might be good for this project i am happy to hear about it.
What i wanna do:
I want to make an app where i can open a folder location of training data(images). Then i want to show the images one by one and mark ROI (rects, cirles, polygons) or bounding boxes of objects. The reason is to mark objects in pictures and then same this regions to a file for later use for image processing in matlab or similar apps.
Thanks.
Qt has lots of very good official tutorials
Try playing with example projects. You can start by changing an example projects to suite your needs.
I need to find a C++ library to find the global movement estimation for a project of super resolution of medical images. But I'm not aware of any of them. Does anyone know a library that could help me do this? I've heard of OpenCV but haven't found anything, maybe I wasn't looking at the right thing.
Thank you.
This looks like a good starting point:
http://www.cvpapers.com/rr.html
although I'm surprised to not see VXL on the list:
http://vxl.sourceforge.net
Have fun!