Multisession scan - Tail disappears

classic Classic list List threaded Threaded
4 messages Options
Reply | Threaded
Open this post in threaded view
|

Multisession scan - Tail disappears

nufanDK2
Hi everyone.

I've tried to do a multisession scan of a house (outdoor), with the Tango-app on the Zenfone AR, but as I re-process the scans on my desktop, the end/tail of the pointcloud and the "route" disappears as the following scans get processed. It seems to be doing it on-the-fly, so when one key-point is added to the front, a key-point at the tail disappears along with that section of the point cloud.

Does anyone have any  idea why this is happening? Is it my laptop that has to little memory (8gb)?



Reply | Threaded
Open this post in threaded view
|

Re: Multisession scan - Tail disappears

nufanDK2
Okay, I might just have stumbled upon the answer:

https://github.com/introlab/rtabmap/wiki/Multi-session#show-the-global-map-and-global-graph

Is it simply because I haven't downloaded the global graph?
Reply | Threaded
Open this post in threaded view
|

Re: Multisession scan - Tail disappears

matlabbe
Administrator
Hi,

Yes, after mapping, you can click on Download graph to get the full map. The first part of the map is disappearing because the time to process a frame is higher than time threshold.

You can also set the time threshold to 0, so that no locations are transfered to long-term memory. When selecting the database in Source panel, you should be asked if you want to process all frames, answering yes will set rtabmap input buffer size to 0 (infinite), so that even if rtabmap cannot process the frames faster than input rate, it will keep them in the buffer. You may lower the input rate to avoid increasing too much that buffer when processing time is over acquisition time.

Cheers,
Mathieu
Reply | Threaded
Open this post in threaded view
|

Re: Multisession scan - Tail disappears

nufanDK2
Thank you so much for the response matlabbe. I got to try it last night, and it worked perfectly. And thank you also for the tip about the time threshold, and the explanation about the buffer and long-term memory. Will definitely try adjusting that next.

And THANK YOU for such a great program