Greg
animals brewing
food and drink gardening
general health
history language
music multimedia
opinion photography
politics Stones Road house
technology
Greg's diary
recent entries
Translate this page
Select day in June 2018:
Su Mo Tu We Th Fr Sa
1 2
3 4 5 6 7 8 9
10 11 12 13 14 15 16
17 18 19 20 21 22 23
24 25 26 27 28 29 30
Select month:
2017 Sep Oct Nov Dec
2018 Jan Feb Mar Apr
2018 May Jun Jul Aug
Today's diary entry
Diary index
About this diary
Greg's home page
Greg's photos
Network link stats
Greg's other links
Copyright information
    
Groogle

Sunday, 17 June 2018 Dereel Images for 17 June 2018
Top of page
next day
last day

Investigating Hugin workflow
Topic: photography, technology, opinion Link here

So why does Hugin not create the correct control points, only under FreeBSD? I've established that the same program, based on the same sources, works as expected under Microsoft.

I have a number of hypotheses: there could be some weird configuration issue, there could be a bug in the code that doesn't bite under other systems, but does under FreeBSD, or potentially there could be a bug in FreeBSD itself.

Of those, the source code bug seems most likely. This used to work, and I “haven't changed anything”, but more particularly I've looked for and tried to eliminate configuration issues. And while FreeBSD doubtless has bugs, it's unlikely that any would be so specific as to cause this kind of problem.

On the other hand, there is one big difference with FreeBSD: the C compiler (and related C++ compiler) is clang, while just about everything else uses gcc. So something like an uninitialized variable could uncover this kind of bug.

How do I go about investigating? First I need to understand what I'm doing. I've already established that I need a project file, which is created by pto_gen. But what does it mean? Went out looking, and asked on the developer's list. Thomas Modes, one of the Hugin developers, pointed me at the unlikely documentation: http://hugin.sourceforge.net/docs/nona/nona.txt.

Huh? nona is just one of the programs that use the project file. From the man page:

NAME
       nona - Stitch a panorama image

But that's misleading. nona transforms the projection of individual images and applies masks. It's not clear if it does anything else, but for the final panorama you need enfuse.

Off to look at the documentation, now nearly 4 years old, but Thomas assures me that it's complete and up to date, apparently conforming to libpano13.

Spent some time working on a description of the project file, making some progress. Only later did I note that the original document is really not a complete description of the project file: in particular, neither the c (control point) or v (variable) lines are described. But at least I'm beginning to make progress, and I've started a case study to go with it.


Truth in advertising photography
Topic: food and drink, photography, opinion Link here

Last week ALDI had various kitchen tools on offer, including this one:


This should be Dicer-1.png.  Is it missing?
Image title: Dicer 1          Dimensions:          1280 x 850, 2959 kB
Make a single page with this image Hide this image
Make this image a thumbnail Make thumbnails of all images on this page
Make this image small again Display small version of all images on this page
All images taken on Sunday, 17 June 2018, thumbnails          All images taken on Sunday, 17 June 2018, small
Diary entry for Sunday, 17 June 2018 Complete exposure details

 

The time I have spent chopping onions! Just what I need! Now admittedly, I've tried this kind of thing before, and it hasn't worked. But this one clearly shows an onion being pressed (and not crushed), and nice, clean onion pieces falling below.

The truth is different. I had to press with as much force as I dared; any more would almost certainly have broken something. And what did I get?

 
This should be Dicer-2.jpeg.  Is it missing?
Image title: Dicer 2
Complete exposure details
Dimensions: 600 x 450, 156 kB
Dimensions of original: 5184 x 3888, 3213 kB
Display this image:
thumbnail    hidden   alone on page
Display all images on this page as:
thumbnails    this size
Show for Sunday, 17 June 2018:
thumbnails    small images    diary entry
 
This should be Dicer-3.jpeg.  Is it missing?
Image title: Dicer 3
Complete exposure details
Dimensions: 600 x 450, 142 kB
Dimensions of original: 5184 x 3888, 2930 kB
Display this image:
thumbnail    hidden   alone on page
Display all images on this page as:
thumbnails    this size
Show for Sunday, 17 June 2018:
thumbnails    small images    diary entry

The onion was crushed, and nothing fell into the collection basin; it all hung through the cutter. And when I scraped it off, I had roughly a third of the onion, in large, uneven, crushed chunks:

 
This should be Dicer-6.jpeg.  Is it missing?
Image title: Dicer 6
Complete exposure details
Dimensions: 600 x 450, 139 kB
Dimensions of original: 5184 x 3888, 2909 kB
Display this image:
thumbnail    hidden   alone on page
Display all images on this page as:
thumbnails    this size
Show for Sunday, 17 June 2018:
thumbnails    small images    diary entry
 
This should be Dicer-7.jpeg.  Is it missing?
Image title: Dicer 7
Complete exposure details
Dimensions: 600 x 450, 188 kB
Dimensions of original: 5184 x 3888, 5227 kB
Display this image:
thumbnail    hidden   alone on page
Display all images on this page as:
thumbnails    this size
Show for Sunday, 17 June 2018:
thumbnails    small images    diary entry

How did they get these nice photos? They faked it, of course. But why? Clearly it goes back.

In passing, it's worth noting a basic design flaw:

 
This should be Dicer-5.jpeg.  Is it missing?
Image title: Dicer 5
Complete exposure details
Dimensions: 600 x 450, 180 kB
Dimensions of original: 5184 x 3888, 4031 kB
Display this image:
thumbnail    hidden   alone on page
Display all images on this page as:
thumbnails    this size
Show for Sunday, 17 June 2018:
thumbnails    small images    diary entry

The top is hinged, so the pressure on the vegetables is at an angle. Given the other problems, that's not so serious, but it would have been possible to make it press from above—just remove the hinge, for example.


Monday, 18 June 2018 Dereel Images for 18 June 2018
Top of page
previous day

Many Hibiscus
Topic: gardening, photography, opinion Link here

By coincidence, and despite the cold weather outside, both our indoor Hibiscus bushes are flowering relatively profusely:

 
This should be Hibiscus-2.jpeg.  Is it missing?
Image title: Hibiscus 2
Complete exposure details
Dimensions: 696 x 388, 189 kB
Dimensions of original: 6202 x 3456, 4590 kB
Display this image:
thumbnail    hidden   alone on page
Display all images on this page as:
thumbnails    this size
Show for Monday, 18 June 2018:
thumbnails    small images    diary entry

Is that a good photo? What about this one?

 
This should be Hibiscus-1.jpeg.  Is it missing?
Image title: Hibiscus 1
Complete exposure details
Dimensions: 757 x 356, 196 kB
Dimensions of original: 4585 x 2158, 2538 kB
Display this image:
thumbnail    hidden   alone on page
Display all images on this page as:
thumbnails    this size
Show for Monday, 18 June 2018:
thumbnails    small images    diary entry

They're both the same photo with different projection. The first is the conventional “wide angle” view (rectilinear projection). And it shows why extreme wide angle lenses are so strange. The second is the image unchanged from the lens (fisheye). The big difference is in the walls: in the first image, they look straight, while in the second they look curved. But at what a price! The pots at the bottom of the photo are now oval, and some of the flowers have been really distorted. In the fisheye view, things are more normal.

But there's more. Here's the same thing again with a cylindrical projection:


This should be Hibiscus-cylindrical.jpeg.  Is it missing?
Image title: Hibiscus cylindrical          Dimensions:          811 x 333, 160 kB
Make a single page with this image Hide this image
Make this image a thumbnail Make thumbnails of all images on this page
Display small version of all images on this page
All images taken on Monday, 18 June 2018, thumbnails          All images taken on Monday, 18 June 2018, small
Diary entry for Monday, 18 June 2018 Complete exposure details

 

Sometimes I think that this might be the best all-round projection for extreme wide angles.


More Hugin case study
Topic: photography, technology, opinion Link here

Continued with my investigation of the Hugin project file today, finding more and more details that aren't in what documentation I have. The result was a significant update of the case study. Things I have found so far:

  1. The GUI doesn't use pto_gen, at least not directly. That explains the difference between the messages and the fact that the GUI can't recognize fisheye lenses.
  2. The i lines are much the same, except that the GUI puts a (spurious) yaw value in them, probably the position where it puts them on the screen.
  3. The GUI also adds a lot of (currently undocumented) v (variable) lines, some of which appear to refer to photometric parameters.

But running the control point detector (cpfind) was the most surprising. Here the differences between the GUI and doing it standalone:

=== grog@eureka (/dev/pts/28) /photowork/Hugin-build-eureka 88 -> cpfind -o e-from-house-0-e-from-house-3-cpfound.pto  e-from-house-0-e-from-house-3.pto
Hugin's cpfind 2018.0.0.5abfb4de7961
based on Pan-o-matic by Anael Orlinski

Project contains the following images:
Image 0
  Imagefile: e-from-house-0.tiff
  Remapped : yes
Image 1
  Imagefile: e-from-house-1.tiff
  Remapped : yes
Image 2
  Imagefile: e-from-house-2.tiff
  Remapped : yes
Image 3
  Imagefile: e-from-house-3.tiff
  Remapped : yes

--- Analyze Images ---
i0 : Analyzing image...
Warning: no TIFFTAG_SAMPLEFORMAT or TIFFTAG_DATATYPE, guessing pixeltype 'UINT16'.
Warning: no TIFFTAG_SAMPLEFORMAT or TIFFTAG_DATATYPE, guessing pixeltype 'UINT16'.
i1 : Analyzing image...
Warning: no TIFFTAG_SAMPLEFORMAT or TIFFTAG_DATATYPE, guessing pixeltype 'UINT16'.
Warning: no TIFFTAG_SAMPLEFORMAT or TIFFTAG_DATATYPE, guessing pixeltype 'UINT16'.
i2 : Analyzing image...
Warning: no TIFFTAG_SAMPLEFORMAT or TIFFTAG_DATATYPE, guessing pixeltype 'UINT16'.
Warning: no TIFFTAG_SAMPLEFORMAT or TIFFTAG_DATATYPE, guessing pixeltype 'UINT16'.
i3 : Analyzing image...
Warning: no TIFFTAG_SAMPLEFORMAT or TIFFTAG_DATATYPE, guessing pixeltype 'UINT16'.
Warning: no TIFFTAG_SAMPLEFORMAT or TIFFTAG_DATATYPE, guessing pixeltype 'UINT16'.

--- Find pair-wise matches ---
i0 <> i1 : Found 21 matches
i0 <> i2 : Found 0 matches
i0 <> i3 : Found 23 matches
i1 <> i2 : Found 22 matches
i1 <> i3 : Found 0 matches
i2 <> i3 : Found 20 matches

--- Write Project output ---
Written output to e-from-house-0-e-from-house-3-cpfound.pto

Detection took 24.268 seconds.

Those matches! 23 matches between the first and the last image!

Comparing the .pto files confirmed it: the control points were all there, and I was able to process the panorama further from that point.

That still doesn't identify the bug, but it gives me not just a workaround, but a better way to do things. Currently I do:

  1. Run a script called Hugin (with capital H; the executable is hugin), which links the images to a specific directory so that I don't have to climb trees to find it, using names starting with 00, 01, etc. to match the references in the GUI. It also resizes them if they're different in size.
  2. Run hugin and load the images from a GUI menu.
  3. Run the Align function, which creates the control points and optimizes them, and which I now need to look at in more detail. There's an alternative which might work correctly. This takes some time.
  4. Further processing, including masking, rotating and possibly change in projection.
  5. Stitching, which also takes some time.

I could easily get Hugin to run pto_gen and cpfind from the Hugin script, which would eliminate steps 2 and 3. Potentially I could also stop it from changing white balance and exposure for each image. Yes, I still need to load the .pto file, but that's relatively quick.

But that's one issue. I still need to understand why the current “Align” function only gives me matches between adjacent images.


Which induction cooker, iteration 4711
Topic: food and drink, opinion Link here

On Wednesday at the latest I need to return the ALDI induction cooktop. What should I buy to replace it with? I had been tending to the Bosch PIJ611BB1E, but today I finally found some details about how to install it: it doesn't have the lip at the front like the ALDI cooktop, so it would have to go back at least as far as the current gas cooktop.

What can I do? I could remove the batten running under the work surface. Despite everything that JG King says, it seems to have no structural function. But there's a door stop on them. What effect would that have? There's another one at the bottom:

 
This should be Kitchen-cabinets-2.jpeg.  Is it missing?
Image title: Kitchen cabinets 2
Complete exposure details
Dimensions: 449 x 601, 121 kB
Dimensions of original: 3888 x 5201, 3167 kB
Display this image:
thumbnail    hidden   alone on page
Display all images on this page as:
thumbnails    this size
Show for Tuesday, 19 June 2018:
thumbnails    small images    diary entry

And then there's the issue with the power. I still can't find out what kind of wiring I need. With the current one, I know I can get by with the wiring I have, as long as I don't use too much power. I can also connect one of the two inputs to the oven, which might mean that we don't need any additional ($$$) wiring from the main switchboard. Can I do that with Bosch? They're keeping that information very close to their chests. And I'd be quite upset if I had to spend $1000 more for something that still annoyed me.


This page contains (roughly) yesterday's and today's entries. I have a horror of reverse chronological documents, so all my diary entries are chronological. This page normally contains the last two days, but if I fall behind it may contain more. You can find older entries in the archive. Note that I often update a diary entry a day or two after I write it.     Do you have a comment about something I have written? This is a diary, not a “blog”, and there is deliberately no provision for directly adding comments. But I welcome feedback and try to reply to all messages I receive. See the diary overview for more details. If you do send me a message relating to something I have written, please indicate whether you'd prefer me not to mention your name. Otherwise I'll assume that it's OK to do so.


Greg's home page This month Greg's photos Greg's links

RSS 2.0 Valid XHTML 1.0!