Maintaining code compatibility between OpenCV 2 and OpenCV 3

Question

My library currently uses OpenCV 2. Now, I am trying to compile the library to use OpenCV 3. It seems that some header files were moved and some constants were renamed. For exam

Answer 1

EDIT:- Read Antonio's answer. You can include compatibility header types_c.h(1) and use OpenCV 2 constant names which works in OpenCV 3 also. This is since, even though constant names differ (e.g. INTER_LINEAR vs CV_INTER_LINEAR), constant values remain the same(2). So functions accepting integer const values should behave no different in OpenCV 2 and 3. Same is the case for other constants also. This way you just have to change header paths between OpenCV 2 and 3.

OLD ANSWER:- This is not possible. OpenCV 3 is not completely backward compatible with OpenCV 2. You will have to change enums, header paths etc between OpenCV 2 and 3. See Changelog(3) for sections about backward compatibility.

You could try things like copying headers around, giving numbers instead of constant names, using #ifdefs all over your code - but sooner or later that is going to fail. I suggest maintaining separate copies and have peace of mind. I would rather ignore OpenCV 2 altogether unless I am currently in production.

Answer 2

The OpenCV guys have put up a transition guide online : http://docs.opencv.org/master/db/dfa/tutorial_transition_guide.html

In essence I think the following have changed:-

• ml module-> The existence of StatModel in OpenCV 3 and that being the root for all classifiers.
• features2d and xfeatures2d -> SIFT and a couple of others have moved from being in the core repository to the opencv_contrib repo
• And many others I am not aware of

As for code working with Opencv3 and opencv2, I usually introduce a prepocessor directive which holds a flag:-

#define HAS_OPENCV3 1
#ifdef HAS_OPENCV3
#include <opencv2/core.hpp> //Any OPENCV3 code
#else
#include <opencv2/core/core.hpp> //Any Opencv2 code
#endif

Since the relative portions are eliminated before compilation, it will also compile with only the OpenCV2 or OpenCV3 libraries. But, introduces a lot of redundancy (which can be avoided by some clever coding).

Answer 3

I had the same problem a few months ago, and I have to contraddict Kiran: I got my relatively big project to build both with OpenCv 2 and 3. Being able to build with both 2 and 3 is very important in a project transition phase, especially because OpenCV 3 is not in stable version yet, and yet has amazing optimization features.

Isn't cv::INTER_LINEAR defined both in OpenCV 2 (header imgproc/imgproc.hpp) and 3? You'll have to replace some constant names, add a couple of inclusions, but that won't be too hard.

Regarding header inclusions, there's a way to find an including path compatible both for Opencv2 and 3.

From the link Kiran posted:

In OpenCV 3 we write

#include "opencv2/core.hpp"

instead of

#include "opencv2/core/core.hpp"

The old method should also work.

Update
In the new Release Candidate, it's reported they have improved compatibility between OpenCV 2.4 and 3.x

Improved compatibility with OpenCV 2.4:

• 2.4.11 now includes "opencv2/core.hpp" and other such headers in addition to standard "opencv2/core/core.hpp".
• smart pointers (Ptr<>) can now be created in both 2.4 and 3.0 style (new ClassName(params) vs makePtr(params))
• trained and stored stat models from opencv_ml 2.4 can now be loaded and used by opencv_ml 3.0 as-is.
• the 2.4=>3.0 transition guide has been sketched: http://docs.opencv.org/master/db/dfa/tutorial_transition_guide.html