SYCAMORE GAP TREE

Photogrammetric reconstruction from archival or crowdsourced images has already demonstrated its potential. But it presents some challenges and may not yield sufficiently accurate and comprehensive results, potentially caused by a poor camera network, low image quality, multi-temporal nature of the data and critical viewing angles. 

Boosted by new photogrammetric AI approaches for the extraction of image correspondences and dense 3D reconstruction, we valorized the legacy of the Sycamore Gap tree through its geometric and colorimetric 3D reconstruction for documentation and restitution to the community. The work demonstrate that photogrammetry coupled to deep learning-based local features can handle the various challenges posed by multi-temporal crowdsourced images and multi-modal datasets.

We used crowdsourced images (ca 330) and 4 videos provided by the National Trust (as flying with drones in the National Trust land without permission is prohibited). As deep learning-based local feature we employed SuperPoint+LightGlue which allow to orient ca 40% more images compared to traditional handcrafted methods. 

The legacy of the Sycamore Gap tree is now made available to the community using a web-based viewer where point clouds and images are displayed together.

LINK TO THE WEB VIEWER