Using the warp transform changes the dtype of the data to float64

In particular, it happens during the warp stream process_frame function

This happens even though preserve_range=True is passed to sk-image.transform.warp, which should preserve it according to the doc.

@dgeissbuhler Do you have an idea why this is happening ?

Minimal example (unit_test branch):

import bob.io.stream
import numpy as np
from pkg_resources import resource_filename

f = bob.io.stream.StreamFile(
    "test/data/input_example.h5",
    "config/idiap_face_streams.json",
    resource_filename("bob.ip.stereo", "config/idiap_face_calibration.json"),
)

color = f.stream("color")
warp_color = color.warp(color)

print(color[0].dtype)
print(warp_color[0].dtype)

Output: uint8 float64

Thanks