The articel article is under open access. With the title
-Feature line detection of noisy triangulated CSGbased objects using deep learning
Citate this work with
@inproceedings{denk_feature_2019,
address = {Jesteburg, Germany},
series = {{DfX}},
title = {Feature line detection of noisy triangulated {CSGbased} objects using deep learning},
volume = {30},
url = {https://www.designsociety.org/publication/42428/Feature+line+detection+of+noisy+triangulated+CSGbased+objects+using+deep+learning},
doi = {10.35199/dfx2019.21},
abstract = {Feature lines such as sharp edges are the main characteristic lines of a surface. These lines are suitable as a basis for surface reconstruction and reverse engineering [1]. A supervised deep learning approach based on graph convolutional networks on estimating local feature lines will be introduced in the following. We test this deep learning architecture on two provided data sets of which one covers sharp feature lines and the other arbitrary feature lines based on unnoisy meshed constructive solid geometry (CSG). Furthermore, we use a data balancing strategy by classifying different feature line types. We then compare the selected architecture with classical machine learning models. Finally, we show the detection of these lines on noisy and deformed meshes.},
language = {en},
urldate = {2019-10-02},
booktitle = {Proceedings of the 30th {Symposium} {Design} for {X}},
publisher = {The Design Society},
author = {Denk, Martin and Paetzold, Kristin and Rother, Klemens},
month = sep,
year = {2019},
pages = {239--250},
}
FeatureLineNet is an open source deep learning appraoch for detecting feature lines on triangulated meshed data. This software runs on Windows, OS X and Linux. FeatureLineNet has been developed by Martin Denk and Prof. Dr. Rother Klemens at the University of Applied Science Munich at the institute for material and building research.
This work is part of the project ANGORA, which is supported by the “Federal Ministry for Economic Affairs and Energy“ with the funding indicator ZF4428401BZ7 in the context of the research network IraSME and ZIM. Responsibility for the content of this publication lies with the authors.
- 3D triangulated Meshes
- Detection of edges in noisy objects
- Usuage of a simple graph convolutional neuronal network (GCN)
- Keras implementation
- Keras
- Pandas
- Numpy
python main.py
- Two data bases 'edges', 'primitive' are available
- You can switch between them by choosing the propper Training data base 'feature_db'
- You can run several machine leraning models or choos our GCN network
import os
import pandas as pd
from src import FeatureVisualizer
from src import run_training
#1.------
# First data set only primitives and simple CSG objects
#feature_db = os.path.join('Data', 'PrimitiveData.feat')
#name = 'primitive'
# Second data set primitives simple CSG objects and bevel, rounded ... corners
feature_db = os.path.join('Data', 'EdgeData.feat')
name = 'edge'
#2.------
# Training of the models
do_training = False
machine_learning = False
model_ft, number_of_categories_ft = run_training.get_model(do_training=do_training,
machine_learning=machine_learning,
database_path=feature_db,
name=name,
epochs=1000)
if number_of_categories_ft != 2:
raise ValueError(f'Binary classifciation requies two categories not {number_of_categories_ft}')
#3.------
# View unlabeled data via mayavi
for stl_file in os.listdir(os.path.join('Data', 'UnlabeledData')):
if stl_file.endswith(".stl"):
file_name = stl_file[0:-4]
visualizer = FeatureVisualizer(model_ft, number_of_categories_ft)
feature_data_frame = visualizer.plot(os.path.join('Data', 'UnlabeledData', file_name + '.stl'),
df=pd.read_csv(os.path.join('Data', 'UnlabeledData', file_name + '.ftu')))