Image 1: Drill core from Pine. It is hard to count the annual rings.

Image 2: Drill core from Pine. You often get dirt on the drill core. It sometimes makes it harder to count and will make automatic counting even harder.

Lärdomar

Jag lärde mig mycket, inklusive följande:

Under åren som följt har det funnits mycket intresse för ämnet. Det har också skett några förbättringar i kamerorna på telefonerna och det har också funnits några intressanta publikationer inom området.

Over the last five years, automated tree-ring detection has rapidly evolved from fragile, rule-based image processing toward robust deep-learning approaches. Early progress came from applying convolutional neural networks to ring-boundary segmentation, enabling far greater tolerance to noise, uneven sanding, and variable lighting. A major milestone was the use of instance-segmentation models such as Mask R-CNN to detect annual ring boundaries in both scanner and smartphone images (Kim et al., 2023). Shortly after, fully automated pipelines emerged that combine deep learning with post-processing and export to standard dendrochronology formats (Poláček et al., 2023). These methods were accompanied by open-source implementations, making state-of-the-art ring detection reproducible and accessible (TRG-ImageProcessing on GitHub). More recently, new annotated datasets and semantic-segmentation baselines have been released to address challenging hardwood surfaces and rough samples (Wu et al., 2024). Transformer-based segmentation models have also been applied to high-resolution wood microsections, improving boundary delineation in anatomically complex samples (García-Hidalgo et al., 2024). At the same time, large foundation segmentation models have lowered the barrier to interactive annotation and rapid prototyping (Kirillov et al., 2023). The open-source release of these models has accelerated experimentation across research and applied product development (Segment Anything on GitHub). Together, these advances now make high-precision, automated ring counting from photographs a realistic foundation for field-ready and mobile applications.

Vi har lagt till data för de senaste iPhone-linserna och kamerorna i listorna

Vi ser fram emot att se var allt detta kommer att sluta