Physical · 2026

Point Cloud Memory Mapper

Give it a photo of a memory. Get back that moment as a 3D point cloud.

Status
Shipped
A stone farmhouse with terraces, cypress trees and flowers rendered as a glowing 3D point cloud on a dark background, with controls to switch between the original photo, the point cloud and a split view.

A photo is flat. A memory isn’t. This project tries to give the memory back some depth.

The idea is simple. Take a familiar photo and turn it into something you can move around in, a place you can revisit from a new angle.

But the service starts the moment someone uploads a personal image. Upload, processing, preview, export and deletion are all part of the experience. So people should always see what happens to their photo, and compare the original with the result before they keep it.

  1. Problem

    Photos of places we care about are flat. Can a single image become something spatial, and still feel like the memory?

  2. Research

    Planned: try a range of non-sensitive photos (places, not people) and talk through what people recognise in the result.

  3. Prototype

    The pipeline: estimate depth from one photo, then turn pixels and depth into 3D points you can rotate. [Verify: model, libraries and output format]

  4. In people’s hands

    Next: invite a few people to try their own non-sensitive photos.

  5. What happened

    What I expect to break: reflections, sky, hair, fine branches and overlapping subjects. A technically good point cloud can still fail if it changes the part of the photo that matters.

How I’m approaching it

The Double Diamond, as planned. Plans loop back once real people get involved, so this one does too.

  1. Discover

    Start with the emotional goal: what makes a photo feel like a memory?

  2. Define

    Decide what a meaningful transformation looks like, and what it must never change.

  3. Develop

    Build a basic photo-to-depth-to-points pipeline.

  4. Deliver

    Inspect the results across very different kinds of images.

  5. Develop

    Go back to the model and output decisions wherever results fail.

  6. Deliver

    Refine the controls and the export.

UX practices

  • Transparent transformationThe source photo stays visible beside the result, so changes make sense.
  • Meaningful controlsA few plain controls for depth and density, not raw technical parameters.
  • Privacy-aware defaultsExample images avoid people, processing is explained, and deletion is clear.

Service design

  • End-to-end service mappingUpload, processing, rendering, export and deletion, including what happens when they fail.
  • Expectation managementDepth estimation is an interpretation of a memory, not a perfect reconstruction.

Privacy notice