I have installed and tested successfully with a simple clahe the CUDA opencv python package. I am not able to make the cuda_FarnebackOpticalFlow.calc work.

This is my code:

    gpu_previous = cv2.cuda_GpuMat()
    gpu_current= cv2.cuda_GpuMat()

    gpu_flow = cv2.cuda_FarnebackOpticalFlow.create(pyramid_depth, pyr_scale, False, 
                                                    window_size, iterations, poly_n, poly_sigma, 0)

    # calculate optical flow
    gpu_flow = cv2.cuda_FarnebackOpticalFlow.calc( gpu_previous, gpu_current, gpu_flow, None )
    flow =

I have the following error at execution:
TypeError: descriptor ‘calc’ for ‘cv2.cuda.DenseOpticalFlow’ objects doesn’t apply to a ‘cv2.cuda.GpuMat’ object

From the documentation the first argument is a
virtual void calc (InputArray I0, InputArray I1, InputOutputArray flow, Stream &stream=Stream::Null())=0
Calculates a dense optical flow. More…
InputArray. How can I adapt the cv2.cuda.GpuMat to the expected formal input argument of the calc method?

shouldn’t this be :

flow = cv2.cuda_GpuMat()
gpu_flow.calc( gpu_previous, gpu_current, flow, None )

(gpu_flow is the farneback object instance, not the output result !)

1 Like

Yes, my bad: calc is a method of the gpu_flow object! Thanks for your help.

1 Like