enblend : out of memory
Affects | Status | Importance | Assigned to | Milestone | |
---|---|---|---|---|---|
Hugin |
Fix Released
|
Critical
|
Unassigned |
Bug Description
On a GNU/linux Ubuntu 8.10 system with hugin 7.0, enblend 3.2, I have 91Go of swap; 4Go ram (but 32bit processor)
I try to make a panorama with 70 pictures.
I reduce the output to 4000px width...
I looked at the amount of memory used by enblend and just before crashing, the use of memory rise in one time.
here is the log :
nona -z PACKBITS -r ldr -e 7 -m TIFF_m -o expo_4000_
enfuse -w -o expo_4000_
Loading next image: expo_4000_
enfuse: only one input file given. Enfuse needs two or more overlapping input images in order to do blending calculations. The output will be the same as the input.
Using 6 blending levels
Generating Gaussian pyramid: g0 g1 g2 g3 g4 g5
Generating Laplacian pyramid: l0 l1 l2 l3 l4 l5
Generating Gaussian pyramid: g0 g1 g2 g3 g4 g5
Collapsing Laplacian pyramid: l5 l4 l3 l2 l1 l0
exiftool -overwrite_
Warning: Too many values specified (2 required) for IFD0:WhitePoint - /home/navillot/
1 image files updated
nona -z PACKBITS -r ldr -e 7.1378 -m TIFF_m -o expo_4000_
enfuse -w -o expo_4000_
Loading next image: expo_4000_
enfuse: only one input file given. Enfuse needs two or more overlapping input images in order to do blending calculations. The output will be the same as the input.
Using 6 blending levels
Generating Gaussian pyramid: g0 g1 g2 g3 g4 g5
Generating Laplacian pyramid: l0 l1 l2 l3 l4 l5
Generating Gaussian pyramid: g0 g1 g2 g3 g4 g5
Collapsing Laplacian pyramid: l5 l4 l3 l2 l1 l0
exiftool -overwrite_
1 image files updated
nona -z PACKBITS -r ldr -e 6.93912 -m TIFF_m -o expo_4000_
[Truncated; see "Full description" attachment]
This does look like a memory leak in enblend.
Though I notice that you are using 'exposure blending' output, but hugin has only identified 5 stacks out of 60.
With 'exposure blending' each image is passed through enfuse, which among other things removes any cropped_tiff offsets. This will increase memory usage, if you don't want exposure blending then you will get better results using 'normal' output.