Projects
Course work, challenge entries and implementations written from scratch. Every number below comes from the logs or report committed in the linked repository, and every image is output from that project.
Four training runs on paediatric brain tumor MRI, held identical except for the loss and the optimiser, then read through the curvature of the loss surface rather than the leaderboard.
6.84 mm boundary error
Two specialised models rather than one shared backbone, aimed at putting individual nuclei back into the tissue context that gives them clinical meaning.
0.62 aggregated jaccard
The Segment Anything Model knows what an object is but has never seen a cell nucleus. Adapting it with low rank updates trains one percent of its weights and gets most of the way there.
0.8233 dice
A 3D U-Net trained with five fold cross validation on brain tumor MRI, where the spread of per case scores turned out to carry far more information than the average.
0.6181 whole tumor dice
Standard U-Net against an attention gated one on BraTS 2021, everything else held fixed. The Dice scores come out identical. The boundary error does not.
10 mm hd95, whole tumor
Transfer learning against training from scratch on chest X-rays, where per class recall shows which way the model leans.
82% test accuracy
Written from scratch in MATLAB. A standard median filter blurs everything it touches; a switching one only intervenes where a pixel is actually corrupted.
Restoring blurred images from scratch, then shifting the assumed blur angle by two degrees to measure how much that assumption is worth.
Two transforms, the same images, compression ratios from 2 to 50. A direct look at what a basis choice actually buys you.
Moving the palette of one photograph onto another by matching distributions in a decorrelated colour space.