MSA- Visual SLAM (no lidar, no IMU) on a drone -- extended version

[@marchaubenstock]

Ethan Eade's notes saved my ass in my masters, I have a lot of respect for him.

[@boredroman2652]

shoutout to the 🐐

[@Willy_Tepes]

It seems it is not good at estimating distance and picks up far too few points. How about sending out a laser dot grid and registering intensity of reflected light with the camera, as well as the dispersion of the grid? You can even use a optical bandpass filter to pre-filter your visual data.

[@enriquejimeneztrejo590]

Which would be the difference between MSA SLAM and ORB-SLAM?

[@charliebaby7065]

thats so awesome! very smooth. so how do you establish the depth of your tracking features? through stereo triangulation? i would assume this would be far simpler than motion parallax or even orb (im just a passionate hobbiest, i have no intention of selling software or people's charitable contributions to my understanding of slam and 6dof methodologies. i painted myself into a corner, stubernly chosing to stick with monocular slam and not turning to ai like midas or zoedepth. i think motion paralax from points maybe my best option. oh, why does the spline fan out towards other way points? is it refining its projected path planning?

[@froimv]

Can I ask, what hardware are you running it on?

[@timreha]

I have a project this would be useful for and have questions? Best way to connect with you William? Thx, Tim

[@main-street-autonomy]

Hi Tim, you can reach me at william@ the name of my company.com

[@idiotophobic]

Is that real time computing?

[@charliebaby7065]

i would most definitly assume so

[@user-sr9pe1zt5i]

how does it perform in indoor environment

[@Moahid6]

The github repo please?

[@samre3006]

Pretty cool.