Volume Downsample Unexpected Behavior #1155
Description
After some investigation into our 3D downsample I'm seeing some unexpected behavior regarding centering and scaling.
Here is a small example script I'm running:
import numpy as np
from aspire.volume import Volume
SIZES = [64, 65]
DS_SIZES = [32, 33]
dtype = np.float64
for res in SIZES:
one_hot = np.zeros((res,) * 3, dtype=dtype)
one_hot[res//2, res//2, res//2] = 1
vol = Volume(one_hot)
# Check vol.downsample()
for res_ds in DS_SIZES:
vol_ds = vol.downsample(res_ds)
# Look at center
cent_vals = vol_ds.asnumpy()[0][
res_ds//2 - 1: res_ds//2 + 2,
res_ds//2 - 1: res_ds//2 + 2,
res_ds//2 - 1: res_ds//2 + 2,
]
print(f"res:{res}, res_ds:{res_ds}\n")
print(f"{cent_vals}\n\n")and the output:
res:64, res_ds:32
[[[0. 0. 0. ]
[0. 0. 0. ]
[0. 0. 0. ]]
[[0. 0. 0. ]
[0. 0.125 0. ]
[0. 0. 0. ]]
[[0. 0. 0. ]
[0. 0. 0. ]
[0. 0. 0. ]]]
res:64, res_ds:33
[[[-0.00132346 0.00395838 0.00395838]
[ 0.00395838 -0.0118393 -0.0118393 ]
[ 0.00395838 -0.0118393 -0.0118393 ]]
[[ 0.00395838 -0.0118393 -0.0118393 ]
[-0.0118393 0.03541068 0.03541068]
[-0.0118393 0.03541068 0.03541068]]
[[ 0.00395838 -0.0118393 -0.0118393 ]
[-0.0118393 0.03541068 0.03541068]
[-0.0118393 0.03541068 0.03541068]]]
res:65, res_ds:32
[[[ 0.00299453 0.0093219 -0.00185942]
[ 0.0093219 0.02873575 -0.00567675]
[-0.00185942 -0.00567675 0.00111061]]
[[ 0.0093219 0.02873575 -0.00567675]
[ 0.02873575 0.08772943 -0.01716358]
[-0.00567675 -0.01716358 0.0033247 ]]
[[-0.00185942 -0.00567675 0.00111061]
[-0.00567675 -0.01716358 0.0033247 ]
[ 0.00111061 0.0033247 -0.00063736]]]
res:65, res_ds:33
[[[-0.00074854 0.00378095 0.00123554]
[ 0.00378095 -0.01909793 -0.00624084]
[ 0.00123554 -0.00624084 -0.00203939]]
[[ 0.00378095 -0.01909793 -0.00624084]
[-0.01909793 0.09646548 0.0315231 ]
[-0.00624084 0.0315231 0.01030116]]
[[ 0.00123554 -0.00624084 -0.00203939]
[-0.00624084 0.0315231 0.01030116]
[-0.00203939 0.01030116 0.00336622]]]
My expectation is that all of the pixel intensity would be in the center pixel and have value 1.