Release from
With TimberVision, AIT presents the world's largest publicly available image data set for AI-supported tree trunk recognition. The system recognises and measures tree trunks precisely - even under challenging environmental conditions - and thus creates the basis for the use of autonomous forestry machines. By providing free access to data sets and algorithms for research and development worldwide, AIT is promoting the digital transformation in forestry.
In forestry, many manual tasks such as stocktaking, timber harvesting and log measurement are not only time-consuming, but also require work in hard-to-reach or hazardous environments. Automated work machines and processes can help here and support the workforce, but also protect them from risks. This requires robust technology that reliably recognises and measures tree trunks and provides the recorded data for further work processes. Until now, there has been a lack of sufficient training and reference data, which is essential for the development and validation of AI-based models. This is where experts Julia Simon, Daniel Steininger, Andreas Trondl and Markus Murschitz from the Centre for Vision, Automation & Control at the AIT Austrian Institute of Technology (AIT) come in. "With TimberVision, the AIT is creating the basis for the next generation of autonomous machines in forestry with an easily accessible system and a unique image data set," explains Markus Murschitz, project manager at the AIT.
TimberVision - the world's largest publicly available image dataset for digitisation in forestry
"Our work presents a novel algorithm for recognising tree trunks in real time, including their geometric properties such as outlines and centre lines. Our approach determines the tree trunks and their components with high accuracy. All the data is merged into a standardised representation,"explains Julia Simon. "The special thing is that our system works reliably even under challenging conditions such as difficult weather conditions or partial occlusion and tracks the tree trunks precisely across image sequences, i.e. recognises them again and again," adds her colleague Daniel Steininger.
More than 51,000 tree trunk components recorded
With TimberVision, they have developed a novel, publicly accessible image data set and an AI model that reliably recognises tree trunks and precisely captures their contours. Over 2,000 annotated colour images and more than 51,000 recorded tree trunk components, including cut and mantle surfaces, make it the largest collection of its kind.
The team recorded the data using standard RGB cameras and annotated it using a specially developed semi-automatic processing pipeline. Several AI models were trained and systematically evaluated for a variety of environmental conditions, different locations, distances, light and weather conditions and tree trunk variations to ensure high model robustness. The accuracy of the model was successfully confirmed in several tests. By combining it with other sensors, the system can be used for automated stocktaking and optimised timber harvesting and loading, for example.
To further advance research, the AIT team is making the entire TimberVision dataset and the developed algorithms publicly available for academic purposes. Scientists worldwide are invited to use and further develop the system.
TimberVision at a glance
- Largest available image dataset for tree trunk detection
- AI-supported algorithms for precise position and contour determination
- Separation and tracking of individual tree trunks across image sequences
- Geometric analysis to calculate centre lines and tree trunk dimensions for more precise handling
- Know-how transfer via GitHub for scientific collaborations
Further information
Open-source image dataset on GitHub: https://github.com/timbervision/timbervision
AIT Large-Scale Robotics Lab: https://www.ait.ac.at/labs/large-scale-robotics-lab