amsterdam-ai-team / urban_pointcloud_processing Goto Github PK

There is a bug in the AHNFuser for ground that sometimes causes large parts of ground to be unlabelled.

If you notice this happening, for now it can be prevented by disabling ground refinement, e.g.
npz_ground_fuser = fusion.AHNFuser(Labels.GROUND, ahn_reader, target='ground', epsilon=0.2, refine_ground=False)

This is due to the fact that a subset of points is queried for AHN interpolation, while the interpolator expects the full set for caching purposes.

A temporary fix is to disable caching in the pipeline:
pipeline.Pipeline(processors=..., ahn_reader=..., caching=False)

I have cloned the repository using PowerShell and installed the required dependencies by running the following commands:

git clone https://github.com/Amsterdam-AI-Team/Urban_PointCloud_Processing.git
cd Urban_PointCloud_Processing
python -m pip install -r requirements.txt

However, I am not familiar with how to proceed further. I am trying to use the notebooks provided in the Usage directory to perform point cloud processing, but I am not sure how to run the code.

For example, the first line of the code in the notebook is import set_path. I am not sure how to set the path to the project directory.

Could someone please provide some guidance on how to set up and run the project, and how to use the notebooks provided in the Usage directory?

Thank you.

I'm a new student, and I had tried a week to run this code but the kernel keeps crashing. It's too difficult for me to cmake the C++ envs. Cloud you please help me with a Docker?

I have already installed the cloudcompare-alpha and cccorelib direct installer on windows. However, it still cannot import cccorelib. Anyone has the same problem?

I just run the complete solution and when I run the block "pipeline.process_file(in_file, out_file)" block, it shows me "module 'laspy' has no attribute 'read'".

What did I do wrong? What should I do?

By the way, please review the following parts of this section. Thanks.

The log is :
INFO - Processing file ../datasets/pointcloud/filtered_2386_9702.laz.
INFO - Caching ground_surface for tile 2386_9702.
INFO - AHN [npz/ground] fuser (label=1, Ground).
INFO - Processor finished in 0.02s, 0 points labelled.
INFO - Noise filter (label=99, Noise).

What are the possible problems? Can you give me some tips so that I can debug?

HI, teacher, I am a beginner of python and segmentation. I was wondering how did you define a road using elevation, like region growing? Here is the function I was confused: 'fusion.BGTRoadFuser(Labels.ROAD, bgt_reader=bgt_road_reader)'.

Thank you so much!

Interpolating the full AHN ground estimate tiles to fill missing data would potentially solve some issues.

Considerations:

• This is always an approximation.
• This will fail for bridges.
• This might fail for quay walls / other water bodies.

I'm trying to run the attached region growing algorithm on jupyter notebook but it keeps saying ''No module named 'cccorelib' '' and pip can't install it !

Bridges are a special class in AHN data, and as such are not seen as ground. This causes them to not be labelled in the point cloud. We should find a way to deal with this.

I'm running region growing on jupyter notebook and it gives error of 'No module named cccorelib'. I tried pip but it didn't work.

This should be:

clusters = self._find_point_cluster(
                    points, c_prime, ground_z + seed_height, max_r=max_r)

where ground_z should probably be a summary / average around the target object.

How can I get a BGT-like datasets of my own country? I'm using OpenStreetMap but I don't know how to get the polygon data like the BGT data.

I just fixed the "laspy" issues but know when I run the block "pipeline.process_file(in_file, out_file)" block again, it shows me

AttributeError: module 'pycc' has no attribute 'PointCoordinateType'

My pycc version is 2.0.0

What should I do? Please help me~ Thank you!!

When processing a single file, the erroe is shown above. By the way, the environment is Ubuntu 20.04 and pycc, cccorelib are built as independent wheels.

Hi,
The requirements file is missing an '=' at scikit-learn

• Store all geodata as .gpkg instead of .csv.
• Modify all code to work with this data directly.
• Will save a whole bunch of conversions between csv, arrays, and shapely polygons.

I tried "pip install cccorelib" but it showed me "ERROR: Could not find a version that satisfies the requirement cccorelib
ERROR: No matching distribution found for cccorelib".
How can I install this module？ Please give me some hint, Thank you very much~~