Thanks for this diagnostic. Question remains : why does this numeric
mismatch only happened on 8 images out of 100 ?
On one of the previous time this had happened, I had tried again the
failed pictures with changed settings in the "levels" ; its default
value for grey was 50% ; I changed it to 52.2. Then most (not all) of
the previously failed pictures got OK. But in the whole, I still get
about the same pourcentage of failed pictures as I got before, maybe
those that fail are not the same.
On 24/05/2018 20:59, johannes hanika wrote:
hi,
this should be fixed in current git master. the levels module in
"auto" mode was counting histogram bins both in float and in int,
resulting in a numeric mismatch and a nan in the pipeline (see output
of darktable -d nan below).
cheers,
jo
[dev_pixelpipe] module `raw black/white point' min: (-0.002037) max:
(0.885251) [thumbnail]
[dev_pixelpipe] module `white balance' min: (-0.003772) max:
(0.885251) [thumbnail]
[dev_pixelpipe] module `highlight reconstruction' min: (-0.003772)
max: (0.885251) [thumbnail]
[dev_pixelpipe] module `demosaic' min: (0.018182; 0.000000; 0.000000)
max: (0.161456; 0.140130; 0.122561) [thumbnail]
[dev_pixelpipe] module `lens correction' min: (0.013752; 0.000000;
0.000000) max: (0.161480; 0.140236; 0.120840) [thumbnail]
[dev_pixelpipe] module `orientation' min: (0.013752; 0.000000;
0.000000) max: (0.161480; 0.140236; 0.120840) [thumbnail]
[dev_pixelpipe] module `base curve' min: (0.012512; 0.000000;
0.000000) max: (0.306351; 0.267746; 0.232010) [thumbnail]
[dev_pixelpipe] module `input color profile' min: (7.238170;
-0.527352; 0.000000) max: (60.891754; 24.719297; 34.346085)
[thumbnail]
[dev_pixelpipe] module `crop and rotate' min: (7.238170; -0.527352;
0.000000) max: (60.891754; 24.719297; 34.346085) [thumbnail]
[dev_pixelpipe] module `equalizer' min: (4.384994; -0.527352;
0.000000) max: (63.679485; 24.719297; 34.346085) [thumbnail]
[dev_pixelpipe] module `levels' min: (5.209786; -0.817160; 0.000000)
max: (99.994942; 34.909893; 51.367264) [thumbnail]
[dev_pixelpipe] module `output color profile' min: (0.147867;
0.000000; 0.000000) max: (1.058848; 0.983380; 0.863795) [thumbnail]
[dev_pixelpipe] module `raw black/white point' min: (-0.002037) max:
(0.885251) [export]
[dev_pixelpipe] module `white balance' min: (-0.003772) max: (0.885251) [export]
[dev_pixelpipe] module `highlight reconstruction' min: (-0.003772)
max: (0.885251) [export]
[dev_pixelpipe] module `demosaic' min: (0.011111; 0.000000; -0.003772)
max: (0.910105; 0.885251; 0.870505) [export]
[dev_pixelpipe] module `lens correction' min: (0.011577; 0.000000;
-0.003549) max: (0.886960; 0.861886; 0.848192) [export]
[dev_pixelpipe] module `orientation' min: (0.011577; 0.000000;
-0.003549) max: (0.886960; 0.861886; 0.848192) [export]
[dev_pixelpipe] module `base curve' min: (0.010574; 0.000000;
0.000000) max: (0.977051; 0.968887; 0.963608) [export]
[dev_pixelpipe] module `input color profile' min: (7.462032;
-15.787125; -9.379805) max: (98.948906; 40.794296; 57.288147) [export]
[dev_pixelpipe] module `crop and rotate' min: (7.462032; -15.787125;
-9.379805) max: (98.948906; 40.794296; 57.288147) [export]
[dev_pixelpipe] module `equalizer' min: (-12.955314; -15.787126;
-9.379805) max: (135.200409; 40.794296; 57.288147) [export]
[dev_pixelpipe] module `levels' outputs NaNs! [export]
[dev_pixelpipe] module `levels' min: (0.000000; 0.000000; 0.000000)
max: (0.000000; 0.000000; 0.000000) [export]
[dev_pixelpipe] module `output color profile' outputs NaNs! [export]
[dev_pixelpipe] module `output color profile' min: (0.000000;
0.000000; 0.000000) max: (0.000000; 0.000000; 0.000000) [export]
[export_job] exported to `/tmp/img_0001_01.jpg'
On Fri, May 25, 2018 at 6:08 AM, johannes hanika <[email protected]> wrote:
hi,
indeed this also comes out black on export on my system, as well as
the preview pipeline seems to remain black.
let me see if i understand what's going on.
cheers,
jo
On Fri, May 25, 2018 at 4:51 AM, Bernard <[email protected]> wrote:
Hi to Everyone,
I have been using Darktable for several years, getting good results for my
purpose. The last version which I used with no known problems, was
v1:1.6.6.0 on my laptop running Ubuntu 13.04.
Now, I have installed Darktable 2.2.5.2~2bpo9+1 on my new desktop (RAM 8 GB,
1GHz processor core 2) running Debian Stretch. On it, same as with the older
versions, I treat Canon CR2s by batch of 100 images. At first, I import 100
images, then I apply a style which I have created quite some time ago,
re-created on my new version, which style flips the images, corrects the
optical defects of my camera, applies the appropriate base curve, sharpens
the image, not with the usually dedicated tool but with the equalizer
instead, sets the levels to 'automatic' with 52,2% grey and 100% black, all
this being applied to the whole batch. Thereafter, I reshape each image
cutting out useless areas. This being done, I export the whole batch to jpg
at 89% accuracy. As said before, this worked perfectly on Darktable
1:1.6.0... but on Darktable 2.2.5.2 it does not work for all images. On a
batch of 100, only 92 images gave good results, the 8 remaining are black.
The good images size about 3,5 to 4,6 MB, that is, somewhat less than what I
obtained before (which was about 1,5 to 2 times bigger), the black failed
images being only abour 150 Kb.
This test has been reproduced about 4 or 5 times over a few weeks, giving
similar results in likewise proportions.
Example of Image numbers that turned black :
IMG_7801.CR2 IMG_7801.CR2.xmp IMG_7801.jpg
IMG_7813.CR2 IMG_7813.CR2.xmp IMG_7813.jpg
Example of Image numbers that turned OK :
IMG_7814.CR2 IMG_7814.CR2.xmp IMG_7814.jpg
IMG_7815.CR2 IMG_7815.CR2.xmp IMG_7815.jpg
you can get them this way:
http://bdebreil.free.fr/Darktable/IMG_7801.CR2
http://bdebreil.free.fr/Darktable/IMG_7801.CR2.xmp
http://bdebreil.free.fr/Darktable/IMG_7801.jpg
etc..
Thanks in advance for your help
Bernard
____________________________________________________________________________
darktable user mailing list
to unsubscribe send a mail to [email protected]
____________________________________________________________________________
darktable user mailing list
to unsubscribe send a mail to [email protected]
