Detection of stem crooks using a CNN trained on synthetic data

Raul de Paula Pires; Henrik Persson; Nils Lindgren; Johan Holmgren

Detection of stem crooks using a CNN trained on synthetic data

Department of Forest Resource Management

Publication: Contribution to conference › Conference abstract (not in scientific journal)

Abstract

Defects like stem crooks significantly impact the value and usability of wood. Obtaining sufficient training data for models that detect such irregularities can be challenging. Thus, data simulation offers a promising solution to overcome data scarcity. By generating annotated datasets, we can train models for detecting defects in standing trees.

Original language	English
Publication status	Published - 2025
Event	SilviLaser 2025, September 29 - October 3, 2025, Québec City, Quebec, Canada - Duration: 1 Jan 2025 → …

Conference

Conference	SilviLaser 2025, September 29 - October 3, 2025, Québec City, Quebec, Canada
Period	2025-01-01 → …

Keywords

Stem crooks
synthetic data
3D simulations
Terrestrial laser scanning
deep learning

Cite this

@conference{d5c4a68d2c7b40239b8ac657b80968ef,

title = "Detection of stem crooks using a CNN trained on synthetic data",

abstract = "Defects like stem crooks significantly impact the value and usability of wood. Obtaining sufficient training data for models that detect such irregularities can be challenging. Thus, data simulation offers a promising solution to overcome data scarcity. By generating annotated datasets, we can train models for detecting defects in standing trees.",

keywords = "Stem crooks, synthetic data, 3D simulations, Terrestrial laser scanning, deep learning, Stem crooks, synthetic data, 3D simulations, Terrestrial laser scanning, deep learning",

author = "\{de Paula Pires\}, Raul and Henrik Persson and Nils Lindgren and Johan Holmgren",

year = "2025",

language = "English",

note = "SilviLaser 2025, September 29 - October 3, 2025, Qu{\'e}bec City, Quebec, Canada ; Conference date: 01-01-2025",

}

TY - CONF

T1 - Detection of stem crooks using a CNN trained on synthetic data

AU - de Paula Pires, Raul

AU - Persson, Henrik

AU - Lindgren, Nils

AU - Holmgren, Johan

PY - 2025

Y1 - 2025

N2 - Defects like stem crooks significantly impact the value and usability of wood. Obtaining sufficient training data for models that detect such irregularities can be challenging. Thus, data simulation offers a promising solution to overcome data scarcity. By generating annotated datasets, we can train models for detecting defects in standing trees.

AB - Defects like stem crooks significantly impact the value and usability of wood. Obtaining sufficient training data for models that detect such irregularities can be challenging. Thus, data simulation offers a promising solution to overcome data scarcity. By generating annotated datasets, we can train models for detecting defects in standing trees.

KW - Stem crooks

KW - synthetic data

KW - 3D simulations

KW - Terrestrial laser scanning

KW - deep learning

KW - Stem crooks

KW - synthetic data

KW - 3D simulations

KW - Terrestrial laser scanning

KW - deep learning

UR - https://res.slu.se/id/publ/145013

UR - https://event.fourwaves.com/silvilaser2025/abstracts/5cb4e13f-04c1-407c-8223-54af598d1742

M3 - Conference abstract (not in scientific journal)

T2 - SilviLaser 2025, September 29 - October 3, 2025, Québec City, Quebec, Canada

Y2 - 1 January 2025

ER -

Detection of stem crooks using a CNN trained on synthetic data

Abstract

Conference

Keywords

Other files and links

Cite this