The question may seem silly to you but is the NodeMicMac project abandoned? There has been no update for 2 years I think.
I tested the tool and I find the rendering excellent. My problem is that with nodeodm my calculation takes 4 hours while the same thing takes 12 hours under NodeMicMac. I have the impression that often only one processor is used (as if mm3d could use only one thread).
Thank you for your answer.
I’m really an amateur and I don’t go very far in the parameters but I didn’t manage to have such a beautiful orthomosaic with NodeODM. However, I repeat, I have no skills and my parameter changes do not exceed orthophoto-resolution or ignore-gsd
That’s fine. We all have a learner’s mind here. If you have the raw dataset you can share as well as the output from NodeMicMac, I can produce something from nodeodm for comparison.
At this time, no ! But python3 is faster than python2 so you could have some enhancements.
In the following weeks, I will try with newer release of MicMac so may be you could expect some improvements …
Hello @ kikislater @ smathermather
I’m coming back because I don’t understand anything anymore.
Indeed I installed NodeMicMac from webodm.sh and when I look at my docker I find dronemapper/node-micmac (last update 3 years ago).
But here I have the impression that you are working on another docker image : opendronemap/nodemicmac
This image is not on docker hub ??? I can’t find it there. I can’t install it from webodm.sh ?
You’re right, docker image was on owner’s repository …
Now I did same thing as nodeodm, putting a github action to push docker images on docker hub. You could find it here : Docker
But I blocked arm64 version due to timeout after 6h of build by default. So if you are on arm, consider this as a transition, and sorry for inconvenience.
Mac chips are now arm, and they’re pretty interesting from a performance to energy efficiency perspective. If you want to avoid building arm for NodeMICMAC, I have no objection, and if someone needs it, they can perform the lift themselves or fund the time to troubleshoot if that is what is needed.
Not to mention that Microsoft is pushing their own (Qualcomm ish) ARM chips harder for consumer units, so… yeah, I think we’ll see more ARM64 “want” short and long term for most everything.
Yeah, the energy savings, chip differentiation/optimization are the current way forward. Intel’s moar and smoller transistors at all costs have gotten too close to the quantum floor, let alone some real manufacturing challenges. And we need to burn fewer electrons for the same work.
But also, if you’re volunteering to do the work, you have a lot of leverage over the things you want to spend your time on.