grasp-pointnet installation issues

[nicolasfrick]

Hi @lianghongzhuo, I experienced some problems when installing this repo:

• The meshpy package depends on an older libboost version. boost/numpy.hpp from Boost.NumPy is deprecated and has been moved to boost/python/numpy.hpp. boost/python/numeric.hpp is also not longer available at (my) version 1.65. Can you please tell which libboost version you use or even update the dependencies?
• Modified dex-net depends on Boost.NumPy as well and there are numerous other import warnings when I run generate-dataset-canny.py. I saw that you removed several files from the original repository, eg. requirements.txt. Can I safely ignore the warnings or will cause this problems with the generated dataset? Below I post the relevant output.

Thank you very much!

nic@salocin:~/code$ /usr/bin/python3 /home/nic/code/grasp-pointnet/dex-net/apps/generate-dataset-canny.py WARNING:root:autolab_core not installed as catkin package, RigidTransform ros methods will be unavailable Unable to import meshrender shared library! Rendering will not work. Likely due to missing Boost.Numpy Boost.Numpy can be installed following the instructions in https://github.com/ndarray/Boost.NumPy WARNING:root:Unable to import pylibfreenect2. Python-only Kinect driver may not work properly. WARNING:root:Unable to import Primsense sensor modules! Likely due to missing OpenNI2. WARNING:root:Failed to import ROS in phoxi_sensor.py. PhoXiSensor functionality unavailable. WARNING:root:Unable to import generic sensor modules!. WARNING:root:Failed to import gqcnn! Grasp2D functions will not be available. WARNING:root:Failed to import OpenRAVE WARNING:dexnet.grasping.grasp_sampler:Failed to import OpenRAVE WARNING:dexnet.grasping.grasp_sampler:Failed to import rospy, you can't grasp now. WARNING:dexnet.api:Failed to import DexNetVisualizer3D, visualization methods will be unavailable All job done. a worker of task 002_master_chef_can start ...

[lianghongzhuo]

We plan to remove some of the dependencies. And right now, you can ignore the warnings, if the code runs. And we will update the requirement.txt then.

[lianghongzhuo]

I hope this commit can let you feel less confused. https://github.com/lianghongzhuo/PointNetGPD/commit/095866183db1f1180db529306ee1be54601c7515

[nicolasfrick]

I have installed the repository from scratch now. No more include errors. Many thanks for your quick response!

Currently I'm running the generate-dataset-canny.py script and the warnings about failed imports popped up but the code runs. I'll try to generate a dataset for 002_master_chef_can object only. I use the preset google_512k data and process it on my laptop. Last time I aborted the processing because it ran for > 15 hours (due to homeoffice I only use a Intel Core i5-4200M processor with two cores and 4 threads). Can you confirm those computing times for a single object?

[lianghongzhuo]

Did you enable multiprocessing? https://github.com/lianghongzhuo/PointNetGPD/blob/master/dex-net/apps/generate-dataset-canny.py#L157 Be careful not to set this number too large.

[nicolasfrick]

The pool_size variable is 1 at my installation. I have used the script with its initial values. But the do_job function releases a lot of threads I guess. The 4 threads at my CPU are continuously at 100% utilization.