Clarification request for data_prepare_semantickitti.py

[sgasperini]

Hi, thank you for your great work, really good paper!

I have an issue with data_prepare_semantickitti.py: when I run it, I get the following error: AttributeError: module 'cpp_wrappers.cpp_subsampling.grid_subsampling' has no attribute 'compute' I did run sh compile_op.sh and installed the required libraries.

• Do you have any suggestion?

I see the preparation file subsamples the points according to a grid, then creates a KDTree with the remaining ones. However, I do not understand the reasons behind this preparation, as it is not mentioned in the paper. Is it an organized way to reduce the amount of points, which then helps towards same-sized samples in a batch?

• Could you please explain it?

Thank you!

[QingyongHu]

Hi @sgasperini, thanks for your interest in our work!

For the first question, I would suggest you check the error message during the compiling.

For the second question, you are right. This is mainly used to reduce the amount of the point clouds, and also reduce the density of the point clouds (a way to increase receptive field). The reason for building the KDTree for each point cloud is to generate regular data for parallel training. We use tf.dataset to generate data flow during training and testing, please refer to this issue #11 for how to generate the data flow.

Hope this would be helpful！

[sgasperini]

Hi @QingyongHu thank you for your quick reply!

I understand the second answer, I will have to look in detail to the usage of the KDTree once I get it running.

Regarding the compilation with compile_op.sh, I actually get no errors. In the image below you can see the created lib and build directories. Do you have any further suggestions on this?

[QingyongHu]

Hi @sgasperini. Unfortunately, I can't give you an exact solution based on this information. My suggestion is to create a virtual environment similar to ours. We use python3.5 and tensorflow-1.11, and the gcc version is 5.4.0.

[sgasperini]

Hi @QingyongHu, ok thank you, I will try that way. Stay safe!

[hikaru-nara]

I encounter the same issue. Have you solved this one? @sgasperini

[hikaru-nara]

I guess the problem is with tensorflow. I first install tensorflow-1.11 with pip but encounter the problem above. Then I tried with conda install and conda said tensorflow-1.11 is not compatible with python3.5. So I upgrade python to 3.6. After upgrading I managed to conda install tensorflow=1.11. pip install -r helper_requirements.txt also brings up problem with version of setuptools in my case. I follow the instructions of pip and it finally work out. But I have to ask, am I the only one who's tensorflow-1.11 isn't compatible with python3.5?? No one has pulled the issue...

[sgasperini]

Hi @hikaru-nara, I'm not sure about the TensorFlow issue you mentioned. For me the problem was that python couldn't find those cpp libraries, even though they compiled correctly. Turned out those libraries were linked to a different python version/instance. This was fixed by running the build commands contained in compile_wrappers.sh individually with the proper python version/instance. Hope this helps!

[hikaru-nara]

@sgasperini After all I've solved the issue. Thanks for the help.

[yfq512]

I guess the problem is with tensorflow. I first install tensorflow-1.11 with pip but encounter the problem above. Then I tried with conda install and conda said tensorflow-1.11 is not compatible with python3.5. So I upgrade python to 3.6. After upgrading I managed to conda install tensorflow=1.11. pip install -r helper_requirements.txt also brings up problem with version of setuptools in my case. I follow the instructions of pip and it finally work out. But I have to ask, am I the only one who's tensorflow-1.11 isn't compatible with python3.5?? No one has pulled the issue...

那你最后的python版本是3.5 tf是1.11么？