Potential memory leak

[jankrepl]

Describe the bug When computing image similarities the memory is not cleared after computation.

To Reproduce Run the below minimal example and have htop running at the same time.

import ants
import numpy as np

img = np.random.random((400, 400))
img_ants = ants.image_clone(ants.from_numpy(img), pixeltype='float')

for i in range(50000000000):
    ants.image_similarity(img_ants, img_ants, metric_type='Correlation')

Expected behavior After each computation of the similarity metric the memory should be cleared. However, the higher the iteration the more memory is taken (eventually leading to out of memory errors).

Desktop

• OS: Ubuntu 18.04.2 LTS"
• Version: 0.1.7 (installed via your wheel in releases)

[stnava]

ok - I verified this. also noting:

• gc.collect() within the loop does not help.

• ending the loop does clean it up.

• same thing with MeanSquares

• does not happen in ANTsR

• same thing in OS X - but ending the loop does not clean up the memory.

• ants.registration does not appear to have the same (degree of?) issues ....

anyway - thanks for this - would appreciate any help with further investigation.

[stnava]

Fixed.

[hassaanmahmood]

I am running code to create atlas, I got full disk error. Is there any solution to this?

 `         for i in range(10):
       Nlin_reg_avg=0

for affines in SUBJECTS[:200]:

    moving_img = ants.image_read( affines )  
    mytx = ants.registration(fixed=fixed_affine_atlas , moving=moving_img, type_of_transform='SyNOnly', grad_step=0.25 ,      reg_iterations=(30,50,20 ),flow_sigma=9,total_sigma=0.2)
    warped_moving = mytx['warpedmovout']
    Nlin_reg_avg=Nlin_reg_avg+(warped_moving.numpy())

Nlin_reg_avg_mean=Nlin_reg_avg/200
Nlin_reg_avg_mean_norm=Nlin_reg_avg_mean/np.max(Nlin_reg_avg_mean)
img = nib.Nifti1Image(Nlin_reg_avg_mean_norm, ref.affine)
nib.save(img, '/Nlin_reg_avg_mean_norm_'+str(i)+'.nii' )
fixed_affine_atlas = ants.from_numpy(Nlin_reg_avg_mean_norm, spacing=fixed_norm_atlas.spacing,
                              origin=fixed_norm_atlas.origin, direction=fixed_norm_atlas.direction)

` My hard disk got full, on investigation I find /tmp folder of 200G with files like these as below: e.g /tmp/tmp1y8z1uo80GenericAffine.mat /tmp/tmp_s4mi6ya1InverseWarp.nii.gz /tmp/tmpeyew06150GenericAffine.mat /tmp/tmpr2o5pl5q0GenericAffine.mat /tmp/tmpuzkont371InverseWarp.nii.gz /tmp/tmp0p486fke1InverseWarp.nii.gz /tmp/tmpwegb539c0GenericAffine.mat /tmp/tmpdctyw2ph1Warp.nii.gz /tmp/tmp7pzvk9g01InverseWarp.nii.gz /tmp/tmpu49tvvao1Warp.nii.gz /tmp/tmphs2lqlh_0GenericAffine.mat /tmp/tmpsu8n32ml1Warp.nii.gz /tmp/tmpzna7lwiy0GenericAffine.mat /tmp/tmptkfaptjx0GenericAffine.mat /tmp/tmphsql8njm1InverseWarp.nii.gz /tmp/tmpxxp5jxid1Warp.nii.gz /tmp/tmp423nwa4l1InverseWarp.nii.gz /tmp/tmp7c30zdvb1InverseWarp.nii.gz /tmp/.Test-unix /tmp/tmp5iyh4mce0GenericAffine.mat /tmp/.font-unix /tmp/tmp7zr5nmc91Warp.nii.gz /tmp/tmpscurinxj0GenericAffine.mat /tmp/tmp0p486fke1Warp.nii.gz /tmp/tmpl22reg7q1InverseWarp.nii.gz /tmp/tmp6olffh461InverseWarp.nii.gz /tmp/tmpqip8xo3b1InverseWarp.nii.gz /tmp/tmppcgfw04u.py /tmp/tmp2303poln0GenericAffine.mat

[ntustison]

Use the outprefix option in ants.registration to save to a known directory and delete the files when no longer needed.

[Jose-Verdu-Diaz]

Fixed.

@stnava I'm having a similar issue when performing image registration on batches of images. Memory usage builds up until running out. Could you expand on how this was fixed, or what was the actual problem?