Nikkilä Colour Analysis

Client: Municipality of Sipoo, Finland, with partner Livady
Location: Nikkilä, Sipoo, Finland
Year: 2020
Sector: Computer Vision · Heritage · Colour Planning

Extracting Nikkilä’s actual palette of building colours from hundreds of geotagged street photos, using computer vision — turning a subjective design question into area-specific, RAL-coded planning guidance.

Following the heritage building survey, Sipoo needed something more specific: concrete colour guidance for how buildings in different parts of Nikkilä should be repaired, extended, or newly built, grounded in what was actually there rather than a designer’s taste.

Built directly on the earlier heritage/typomorphology survey of Nikkilä, refining and applying its area boundaries to a new computer-vision colour analysis — the third phase of a continuing engagement with the Municipality of Sipoo and partner Livady. The method reused and refined an approach SPIN Unit first developed for a 2019 colour analysis of Narva’s wooden-house district in Estonia.

Roughly 360 geocoded photographs were taken across Nikkilä’s built and natural environment. A three-stage computer-vision pipeline segmented natural from built elements, further separated facades, streets, and details, then extracted each element’s dominant colours — mapped back onto the city using each photo’s geolocation. Recommendations were split by intervention type: repair of existing historic buildings, context-respecting infill, and significant new construction.

Delivered as a 16-page branded report with a condensed client-facing version, covering four sub-areas of Nikkilä, each with recommended RAL Classic and RAL Effect colour codes.

  • Colour-zone boundaries were anchored to the earlier heritage survey, not drawn arbitrarily — colour policy follows preservation value and building typology, not just geography.
  • Recommendations are genuinely differentiated by what’s being built: repair of historic buildings, contextual infill, and bold new construction each get distinct palettes within the same zone.
  • The village/manor landscape area was steered toward muted red and yellow ochre with unpainted or tarred timber; the former hospital/heritage area toward greys and dark reds reflecting its early-20th-century brick materials; new-build zones near the town centre toward bolder greens, oranges, and turquoise.
  • The team openly reported the method’s limitations — inconsistent weather and lighting during the photo survey, and shadows reducing colour accuracy — and proposed a portable colour-meter field survey as a next-step improvement rather than overselling the photo-based method.
  • One area, the Jokilaakso river valley, was deliberately excluded because it already had its own dedicated colour plan — disciplined scoping rather than claiming blanket coverage.

Gave Sipoo’s planners concrete, actionable colour guidance directly usable in building-permit decisions — a rare example of computer vision applied to a genuinely aesthetic planning question, delivered by a three-person team.

Download