An imbalance handling method to improve the detection and classification of symbols depicted in Piping and Instrumentation Diagrams (P&IDs)
See demo.ipynb for a jupyter notebook demo on how to use this method.
The original symbol dataset can be downloaded from here: https://www.dropbox.com/s/sj277k4slmrv3qc/symbols_combined_pixel_red.csv?dl=0
The generated symbol dataset, once CDSMOTE has been applied, can be downloaded from here: https://www.dropbox.com/s/ll562q3gjqyrhp9/cdsmotedb_symbols_combined_pixel_kmeans.csv?dl=0
Please reference this method as follows:
-
Elyan E, Moreno-García CF, Johnston P. Symbols in Engineering Drawings (SiED): An Imbalanced Dataset Benchmarked by Convolutional Neural Networks. In: Engineering Applications of Neural Networks (EANN). 2020. p. 215–24. https://doi.org/10.1007/978-3-030-48791-1_16.
-
Jamieson, L. Moreno-García CF, Elyan E,. A Multiclass Imbalanced Dataset Classification of Symbols from Piping and Instrumentation Diagrams. In: International Conference on Document Analysis and Recognition (ICDAR). 2024.
or use the BibTex entries below:
@inproceedings{Elyan2020, author = {Elyan, Eyad and Moreno-Garc{'{i}}a, Carlos Francisco and Johnston, Pamela}, booktitle = {Engineering Applications of Neural Networks (EANN)}, doi = {10.1007/978-3-030-48791-1}, isbn = {9783030487911}, keywords = {Imbalance Datasets Multiclass,P{&}ID,classification,cnn,engineering drawings,id,imbalanced dataset,multiclass,p}, pages = {215--224}, title = {{Symbols in Engineering Drawings (SiED): An Imbalanced Dataset Benchmarked by Convolutional Neural Networks}}, year = {2020} }