I'm currently undergoing a batch test with two feature extraction
methods (SIFT/MOPS). There were some strange results in the past and I
want to see what is happening.
1. SIFT: The results are repeatable (batchmode on/off)
2. SIFT: Same image pairs (switching source/target) -> other results
3. SIFT: a) Some pairs have no matching features after batch-processing
b) some no feat. pairs, have matching feat. doing them manually
4. MOPS: causes exceptions and won't run via macro (at least here)
Can anyone confirm these? Especially 2 and 3 considering SIFT are
strange and not that nice for most of the use-cases I can imagine.
Another curiosity can be reproduced like this.
1. download above zip-file
2. run macro with "a_source" as source and "b_target" as target-folder
3. The results of these ten pairs should be 0 for each (our results)
4. checking any pair manual returns the same, 0 matching features
5. restarting Fiji and repeating step 4 is different (at least here)
All ten images per folder are identical. These two folders were
populated by copies of "uvn_0079495.jpg" (source folder) and
"uvn_0079496.jpg" (target folder).
Any thoughts and help in understanding the/my problem is much appreciated.