[CREATE] [hugin-ptx] Re: Lens correction database

[Pablo d'Angelo]

Bruno Postle schrieb:
> On Thu 17-May-2007 at 15:02 +0200, Pablo d'Angelo wrote:
> 
>>> During the stitching process, hugin evaluates the quality of the 
>>> optimisation by calculating error distances, why not improve 
>>> this with some extra checks and give the user an option to 
>>> upload the results from _all_ good optimisation passes.
> 
>> This could happen through uploading an anonymized pto file (image 
>> filenames etc. removed) to some central server.
> 
> The pto files can get very large, especially when calibrating lenses 
> with thousands of control points.  Is it practical to upload a 30kB 
> pto file with a dial-up connection?  

If I had a dialup only (common for some people in germany, DSL is not 
available everywhere), I wouldn't want to used for uploading such kind of 
data anyway. Maybe some caching scheme might help.

 > How many is it practical to receive concurrently?

Hmm, maybe a binary format would be more efficient than the pto text files. 
Except for extremely large projects, the size required for transmission 
would be a smaller than 10 kB (uncompressed). But still, it could lead to 
quite some burden on the receiving server, I agree.

> Perhaps this is too simplistic: but the lens parameters could be 
> submitted to a server with a single http GET just by appending them 
> to the URL:
> 
> http://example.com/lens?CameraModel=Nikon%20D100;CropFactor=1.5;a=0.01;b=-0.021;etc...
> 
> This could be received by a CGI or PHP script and stuffed into an 
> SQL database, but it would be even simpler to have no script on the  
> server and pull the information out of the server logs - This system 
> is _very_ scalable even with modest server hardware.

Yes, this sounds like a good idea to get a scalable system.

The only drawback is that less information is available, and most processing 
to determine if a stitch is good for calibration purposes needs to be done 
in hugin itself.
Having access to the pto file would provide much more flexibility, and 
improvements in the algorithms to mine the calibrations could be done on the 
existing database.

ciao
   Pablo