replace scikit-learn nearest neighbor with open3d about manipulation HOT 5 CLOSED

Seems like a good idea. I also got the following error while trying to use scikit-learn in Docker:
RuntimeError: module compiled against API version 0xc but this version of numpy is 0xb

which stopped after I did pip3 install numpy --upgrade.

Strange, but it seems like the version installed by sudo apt-get install python3-numpy in install_prereqs.sh is not high enough? Would it be good to put a version requirement for numpy in requirements.txt?

from manipulation.

ugh. i already did something similar for scipy to make CI happy. i'm ok if you need to do it.

from manipulation.

@hjsuh94 -- i've made the change in my pose.ipynb. can you propagate it to the exercises, and let me know? then we can remove the build dependencies, too.

from manipulation.

Got it! I will make a PR after running some checks like uploading to Gradescope.

Meanwhile, I think you have a typo in pose.ipynb.
distance = np.linalg.norm(point_cloud_A[:,i] - point_cloud_B[:,i])

should be
distances[i] = np.linalg.norm(point_cloud_A[:,i] - point_cloud_B[:,indices[i])]?

from manipulation.

oops. gulp. that actually also revealed that i'm not actually using the distances anywhere. So I've simply removed that part. thanks for catching it!

from manipulation.