Not sure if this will reach the intended recipient [edit: originally replied via email], but I'm glad to see

[cmcfadden]

Not sure if this will reach the intended recipient [edit: originally replied via email], but I'm glad to see
that development is still going on with this project. I reached out a while
ago intending (but never getting around) to using this for my LadyBug
scanner project, which uses a 3D printer or other CNC to scan by taking a
regular grid of overlapping images using a USB microscope. Example data on
a PCB here:
https://drive.google.com/file/d/1z4HVS1OW51BIg2poLdHDMqvHnxTrtxkN/view
Would love to see an attempt on it; happy to provide more info and a
plethora of other example sets if needed.

~Ahron

On Sun, Aug 30, 2020 at 10:45 AM Ross Esmond [email protected]
wrote:

It definitely runs. I successfully stitched two stacks of images together
on my side machine after following my instructions precisely, but I only
tested the most basic functionality. If you know of any other specific
features I should test I can hit those as well. I can't anything about
feature extraction or cores.

What I can say is that the SIFT algorithm—the core feature that was
missing with more recent versions of opencv—is available in
opencv-contrib-python. It was hidden in opencv and opencv-python because it
was patented in the US. You could only use it if you compiled opencv with a
"non-free" flag, but one of the versions of opencv accidentally included it
by default. That's the version you were on. Opencv-contrib-python, the
package I switched to, has the "non-free" flag set to true, and so it
builds with the SIFT algorithm by default. That's why I was able to switch
to the most recent version.

By the way, the SIFT patent is twenty-one years old as of March of this
year, so it has expired. The Opencv team has been informed and they're
considering turning the feature on by default.

Its possible anaconda is doing something else that I'm not aware of. If
you know of more issues with pip I can check for those as well.

I can also prepare a lighter rewrite of the README which cleans up dead
links but keeps all the instructions that were there before.

—
You are receiving this because you are subscribed to this thread.
Reply to this email directly, view it on GitHub
#6 (comment),
or unsubscribe
https://github.com/notifications/unsubscribe-auth/AJ74LVY74FULI2KTVVITZBLSDJQYXANCNFSM4QOPZJ5Q
.

Originally posted by @Waynewaynehello in #6 (comment)