search

eth-ait / dip18

Deep Inertial Poser: Learning to Reconstruct Human Pose from Sparse Inertial Measurements in Real Time
GNU General Public License v3.0
146 stars 47 forks source link

Question: Super Noisy Acceleration Data? #19

Closed jyf588 closed 2 years ago

jyf588 commented 3 years ago

Hi!

Thank you so much for creating this really useful IMU dataset. I encountered a problem when I was playing around with it and was wondering if you could kindly help me with it:

When I load the raw IMU data, select a random sensor, and visualize its readings, the IMU orientations look pretty much reasonable and consistent, but the acceleration visualizations (direction of plotted 3D vectors) almost look white noise to me as they wildly change directions every timestep.

Is this something you would expect or I must be doing something wrong? How could I convince myself (e.g. through visualization code) that I am using the IMU acc readings correctly?

Thanks!

xdtcssdi commented 3 years ago

HI, Can we communicate the generation of Transpose dataset.

kaufManu commented 3 years ago

Hi,

@jyf588 It is normal that accelerations look noisy. It is also not so easy to interpret them visually because you might feel like they behave counter-intuitively (e.g. when an arm is moving upward, but it is slowing down, then the acceleration will point in the opposite direction of movement). Also make sure that the accelerations you are looking at do not contain the gravity component.

One way of sanity-checking the accelerations is to synthesize them via finite differences by using a vertex on the SMPL mesh and then checking whether those synthetic accelerations match the real ones. A perfect match is unlikely, but at least it should give you a rough idea. With some smoothing to remove the noise you should get a reasonable match.

As @xdtcssdi mentions, feel free to take a look at their repository (https://github.com/Xinyu-Yi/TransPose) which might be helpful, too.

Best regards Manuel