Room 04 / 06 / The House
Lookout
Computer vision: reading a match from one camera and a forest through cloud.
The Lookout
This is where I work with pictures: reading what a camera or a satellite sees and turning it into facts.
My dissertation took ordinary television footage of a football match and worked out where every player was and what happened. At a hackathon in 2026, my team found cleared forest from radar, which sees through cloud.
I learned that the model is only one step. Most of the work is checking that the numbers it gives are true.
the telescope on the veranda / 2026
Football Intelligence
The dissertation. From one broadcast camera, it follows every player, maps them onto a top-down pitch, labels what happens and writes commentary from the numbers.
The right label stays on the right player about 92% of the time.
Read the chapterthe radar map / 2026
Deforestation Detection
Clouds hide forest from ordinary satellite photos. Radar passes through them. Eight years of radar over 100,000 ha showed where 6,545 ha of forest had gone.
About 65 square kilometres of cleared forest, roughly 9,000 football pitches.
Read the chapterthe prize
The hardest track
Treefera's hackathon in June 2026 had 25 teams and several challenge tracks. The radar project won the hardest one.
the pitch sketch pinned by the window
Where the dissertation started
An earlier football pipeline came first: YOLOv8 to find players, ByteTrack to follow them, team colours grouped by a model, and a map of the pitch from above.
