opencv: Incorrect error message for non-integer points in rectangle draw function: TypeError: function takes exactly 4 arguments (2 given)

I got the same issue. I resolved it by converting all the coordinate of the bounding box into int instead of float. You are facing the problem because of one coordinate in float

@MLDLCo The issue isn’t that ints are required, rather that the wrong error message is being given for this problem. It should give TypeError: integer argument expected, got float. instead of TypeError: function takes exactly 4 arguments (2 given).

The root cause of the issue is current implementation of the overload resolution in the generated Python bindings. First of all it tries to parse arguments for

void rectangle(InputOutputArray img, Point pt1, Point pt2, const Scalar& color, 
               int thickness = 1, int lineType = LINE_8, int shift = 0);

but it fails (because floats are not convertible to integers as was already stated above) - it sets error message, but in C++ interface there is an overload, that clears previously set error message:

void rectangle(InputOutputArray img, Rect rec, const Scalar& color,
               int thickness = 1, int lineType = LINE_8, int shift = 0);

that accepts tuple with 4 elements as the second argument. First it is checking the tuple length - you provide tuple with 2 elements instead of 4 - error message as reported to user as the last set.

In my opinion, to solve the problem with inappropriate/misleading message the following steps should be done:

1. Update the logic for argument parsing - argument name or at least argument index should present in the reported error (if possible, e.g. while parsing tuples with default PyArg_ParseTuple it is not possible).
2. Collect all error messages during overload resolution.
3. If the last overload fits - use it and forget about previous errors.
4. Otherwise raise custom OpenCV exception of OverloadError type with previously collected messages.

The main drawback of the suggested way is a small overhead increasing for functions calls.

This is cause because you tuples contain floats. OpenCV takes integers within those tuples.

This could be a integer overrun.

i recommande to install the version python 3.4.8.29 pip install -U opencv-python==3.4.8.29

OpenCV should get more tolerant of various data types when it expects exactly tuples of integers. it would be great if OpenCV accepted floats in addition to ints, and if it accepted numpy arrays in addition to tuples. it would be great if that was part of the bindings generation so all OpenCV functions benefit from more tolerance in accepting arguments. drawing functions (taking a shift argument) would benefit from a special case of accepting floats.

I had the same error which ended up being caused by my color tuple being a tuple of numpy.int32 values which I thought were allowed in cv2. I had to manually convert the items to the stock python int() type.

This error message definitely needs to be fixed since it leads you down the wrong rabbit hole for a long time before you realize what the actual problem is.

I got the same issue. I resolved it by converting all the coordinate of the bounding box into int instead of float. You are facing the problem because of one coordinate in float

@MLDLCo Thank You.

@MLDLCo thank you very much! Your solution solves my problem.

The reason of this issue is I set two locations as float type, but I think this is quite confusing 😅

在 2019-10-14 14:13:07，Ajith notifications@github.com 写道：

i think it is because of the reason that you are executing a python2 code in python3 but i am not sure of it

— You are receiving this because you commented. Reply to this email directly, view it on GitHub, or unsubscribe.