Hydrants@Charles River in 35mm Films
An Object We All Trained On
Before Cats, There Was Fire Hydrant
oid/m/01pns0 & coco-11: Portraits
Shot on 35mm film near Harvard, this series revisits a forgotten icon of early computer vision — not as data, but as presence.
Once, the fire hydrant stood at the frontier of machine perception: a geometric constant, bold in form, universally recognizable. These red, yellow, rusted sentinels taught early models to see — not with imagination, but with recognition.
This project turns the lens around.
Wandering the streets by the Charles River, I photographed fire hydrants on 35mm film, embracing the slow cadence of analog. Each hydrant reveals its own posture, patina, and context — a quiet architecture shaped by both utility and neglect.
What was once a bounding box is now a portrait.
README.md– You’re reading it/images/– Digitized 35mm scans of 36 hydrants, in JPEG/metadata/– To be added after the second roll finds its way through the camera :)
Fire hydrants are among the most frequently labeled objects in early object detection datasets like COCO (category ID 11) and Open Images (class /m/01pns0). (Do you know why early vision folks used hydrants for camera calibration?)
With their bold shapes, fixed positions, and street-corner ubiquity, they offered the perfect subject for machine vision’s formative years — simple, reliable, and everywhere. Before deep networks turned to cats, faces, or dreams, the hydrant stood as a quiet constant.
This project revisits that role — not as annotation, but as presence; not as data, but as infrastructure made visible.
- Camera: Canon AE-1
- Lens: Sigma 35-70mm F2.8-4
- Film: FUJIFILM 400 Color Negative
- Scanner: idk, Hunt's Photo did it for me
- Location: Cambridge, MA (near Harvard & Charles River)
All images and text are © 2025 Leslie Gu.
For non-commercial use only. Contact for other usage.
