Forward facing camera, LSD-SLAM about ethzasl_msf HOT 11 CLOSED

Ok thanks, I found a mistake in one of my nodes - that seems to work now... Great! I'll start writing up a wiki in the next days on how to set up LSD-SLAM with the MSF!

from ethzasl_msf.

Please see issue #83

from ethzasl_msf.

And regarding the pose output, you might need to override the subscriber detection. I had problems with this too, even after initialising the filter.

from ethzasl_msf.

Also, looks like you copied the params from my SVO wiki tutorial. Can you add the filter init how-to to the "running" section of the wiki please.

from ethzasl_msf.

I added the init how to to the wiki, should I add the subscriber thing, too, or will that be fixed in a future update of the framework?

from ethzasl_msf.

Ok, so with the rotation quaternion from #83 I don't really get any better results. I still get the warning:

[ WARN] [1419366341.634866394]: Wanted to insert a value to the sorted container at time 1419366341.083073139 but the map already contained a value at this time. discarding.

Does anyone know where this comes from?

from ethzasl_msf.

@MarcGyongyosi The warning you get results from duplicate measurements either from your visual module or your IMU. The message says simply that a given measurement already exists in the filter at a certain time and thus is discarded. Since we have double resolution on the time and thus measurements are unlikely to collide as such, it is probably the case that your time-stamping is not correct.

from ethzasl_msf.

@MarcGyongyosi So can you get proper estimated metric scale in the EKF output? I had run the same experiment sometime before LSD-SLAM's public release, but didn't get an opportunity to fly it as I didn't have a good-enough onboard companion at that time. I wasn't able to run LSD-SLAM satisfactorily on a Odroid U3.
I would strongly advise getting a flight in. Please drop me a note if you do, and please cite the computer you're using for flights.

from ethzasl_msf.

So far I don't have the proper setup to do an actual evaluation, but from the output I get on my Macbook Air with an i7 it seems to be doing just fine (I'm on vacation and only have that with me). Also, the metric scale seems to be somewhat accurate, though I have no quantitive data on that.

I will have an NVIDIA Jetson TK1 onboard a completely custom quadcopter. As soon as I get to the point where we have our actual prototype, I will start collecting flight data.

from ethzasl_msf.

Any news on that wiki? trying the same thing at the moment and i'm a bit lost.

from ethzasl_msf.

@MarcGyongyosi, Hello! Any news on that wiki? I'd like to run LSD SLAM and fuse visual odometry from it with IMU to estimate scale.
Thanks

from ethzasl_msf.