Elaborate on Usage

[tonyf]

I was wondering if you could elaborate on the usage. I'm trying to compute the optical flow between two images.

When passing the numpy arrays for the images to calc_flow as follows: result = optflow.calc_flow(im1, im2) I get the following error: *** TypeError: unorderable types: NoneType() < int()

Thanks!

[theHamsta]

Hi there!

It was actually just a broken assert with the default values. You had to set the parameter quit_at_level to some value. Could you check again with the updated version.

Maybe you can report a minimal example to reproduce in case it still fails.

The following test should now work:

def test_default_values():

    a = np.ones([20] * 3)
    b = np.ones([20] * 3)

    optflow = farneback3d.Farneback()

    rtn = optflow.calc_flow(a, b)
    assert rtn is not None

[tonyf]

Ah! Thanks for the update. It solved that issue.

Just a heads up the code you posted results in a failed assertion as the bumpy array has type int rather than np.float32. You might want to note that somewhere (implicitly casting would lead to confusion).

Running my code I now run into another error: pycuda._driver.LogicError: cuMemAlloc failed: invalid device context

Stack trace: File "farneback_flow.py", line 22, in run return self.optflow.calc_flow(batch[0], batch[1]) File ".../python3.5/site-packages/farneback3d/_farneback3d.py", line 124, in calc_flow imgs = [gpuarray.to_gpu(next_vol), gpuarray.to_gpu(cur_vol)] File ".../python3.5/site-packages/pycuda/gpuarray.py", line 1049, in to_gpu result = GPUArray(ary.shape, ary.dtype, allocator, strides=_compact_strides(ary)) File ".../python3.5/site-packages/pycuda/gpuarray.py", line 210, in __init__ self.gpudata = self.allocator(self.size * self.dtype.itemsize) pycuda._driver.LogicError: cuMemAlloc failed: invalid device context

wherebatch[0] and batch[1] are each (1080, 1920, 3) numpy arrays.

[theHamsta]

Ok, good to know I try to fix that as soon as I have time! :+1: