Talk:OSMCAM
You can use our community sourceforge project to add your softwares to the CVS and publish a build image for OSMCAM. ;)
--Alexandre.Poltorak 15:02, 28 August 2008 (CDT)
Since my current code is based on the work GPL'ed of Rob Hess, I was waiting for his comment on the project. Next to this that could interest other people:
This is what we could be targeting. Never the less, the changes are now in git, can you give me developer status on sourceforge? I guess you can guess my nickname ;) --Skinkie 19:39, 28 August 2008 (CDT)
done --Alexandre.Poltorak 00:38, 29 August 2008 (CDT)
Test recognizer is in, we are now testing on photos. So we will have fun as long we are not selected for developer previews of mighty camera's ;) --Skinkie 21:35, 1 September 2008 (CDT)
hehe.. The camera will be sent normally tomorrow. Could you please describe how to use the script ? --Alexandre.Poltorak 23:00, 1 September 2008 (CDT)
Test script howto
./htdocs ./htdocs/matches ./htdocs/matches/batch1 (both directories store the final result) ./htdocs/matches/batch2 ./htdocs/matches/processall.sh (the script that matches key files from template with uploaded) ./htdocs/matches/preview ./htdocs/templates (directory containing all key files that we want to look for) ./htdocs/templates/fetchall.sh (some script to get all dutch trafficsigns from wikipedia) ./htdocs/uploaded ./htdocs/uploaded/batch1 ./htdocs/uploaded/batch2 ./htdocs/uploaded/batch3 ./htdocs/uploaded/process.sh (the script that generates key files from the uploaded contents)
Our webserver structure is like the above. For every new upload session I have created a new batch directory. The script process.sh with as argument the number of the batch, run siftfeat (key generator) on all the files in that directory.
Process all compares n-templates with m-input, the output it written in a filename structure where the highest prefix show the highest amount of matches.
The binaries this program depends on can be found in the sift directory, siftfeat and match. The most simple example would be:
siftfeat -o <<outputkey>> <<inputimage>> siftfeat -o <<outputkey2>> <<inputimage2>> match <<inputimage>> <<inputimage2>>
Sift works quite well on photo's, but terrible on flat pictograms :)
