Champion's Award and Robot Performance winners at the London Regional. Next stop: UK Nationals in Harrogate.

Top overall trophy at the London Regional. Judged across robot game, robot design, innovation project, and core values.
Highest score on the UNEARTHED mat at the London Regional.
Qualified for the FIRST LEGO League UK Nationals — Harrogate, 25 April 2026.
Much of the Maya world is still buried under jungle canopy. Archaeologists use LiDAR to see through the trees, but labelling a site like Chactún by hand takes months. We trained a model to do it automatically.
IoU on the Chactún test set · Best model: UPerNet + ConvNeXt-Small




Green = platforms · Red = buildings · Blue = aguadas · LiDAR + CHM input
Manual scanning takes months per site. Many are destroyed by development before anyone gets to them.
Semantic segmentation over LiDAR plus canopy-height data, predicting three feature classes in one pass.
Tested 4 architectures and 3 loss functions. Tuned per-class thresholds (0.95 / 0.63 / 0.48) to lift mean IoU to 67.1%.
Training code and a Gradio demo on GitHub, so archaeologists can run the model on their own tiles.
UK Nationals are on 25 April 2026 in Harrogate — the last tournament of the UNEARTHED season.




Seven students from Harrow School competing in the FIRST LEGO League Challenge. We handle the robot, code, innovation project, and everything around them.