SuperGluePretrainedNetwork VS ORB_SLAM3

Code: https://github.com/magicleap/SuperGluePretrainedNetwork

This basically includes a deep learning based approach to do keypoint detection, and match them across image frames. This includes papers like SuperPoint, Superglue, and more. There is also a way to do dense matching with neural networks.

Feature matching is the way to go imo. Try out OpenCV's inbuilt feature matching methods like SIFT and FLANN. If the performance is poor, you can even try out CNN aided matching algos like SuperGlue Link (CVPR2020)

SIFT and brute force matching is your best bet in classical computer vision if you're unconcerned with runtime. There are methods from deep learning that can perform better, somewhat domain dependent. Check out superpoint and superglue from magic leap. https://github.com/magicleap/SuperGluePretrainedNetwork

Thank you for the recommendation. Other's I've talked to have also mentioned slam. I think I might use this repo that looks promising: https://github.com/UZ-SLAMLab/ORB_SLAM3

https://github.com/UZ-SLAMLab/ORB_SLAM3 looks promising. Thank you for the recommendation!

ORB_SLAM3 on GitHub: https://github.com/UZ-SLAMLab/ORB_SLAM3

Yeah pretty much. I'd say it can still be useful if your custom application only needs one or two new modules and can otherwise be formed from existing code. It's great if you make a lot of different systems with similar sub-modules. The bigger research institutions also open source a lot of their code with ROS integrations, so you can just drop in state of the art modules where appropriate. e.g. OrbSlam3,

You could try your luck with OrbSLAM3 or Kimera, they're pretty close to state of the art and open source. Might have a bit of a learning curve though.

I hear good things about stereo-visual + inertial and OrbSlam3 seems pretty hot, but this is a couple years old now and I'm not active enough in the field to give a definitive opinion.

I'm currently in the planning stage of building an indoor autonomous MAV/drone. Up until this point I thought this would require a SLAM solution, such as ORB-SLAM3, but now I see this framework by Intel which doesn't mention SLAM anywhere. All of the literature seems to be about SLAM so maybe I just don't know the terminology for what I'm looking for.