Maintenance work on complex industrial equipment depends heavily on technicians knowing which component sits where, and which one to remove first. That knowledge currently lives in manuals, in the heads of senior technicians, or simply gets relearned through trial and error on site. As equipment ages and experienced staff retire, this tacit knowledge becomes harder to pass on, and field technicians lose time figuring out disassembly sequences that an expert would know instantly. A visual, interactive guide derived directly from the physical device — rather than from static manuals or generic CAD models — could close this gap quickly and cheaply, without requiring the original design files.

Key questions to be answered in the project:
1. What is the minimum photo coverage and capture process needed to reconstruct a 3D model with usable segmentation quality?
2. How accurately can individual components be automatically segmented and labelled from the reconstructed model, and how much manual correction is needed?
3. Can the photogrammetry-to-interactive-guide workflow be made fast enough for practical use by field technicians?
4. What interaction model (touch, voice, AR overlay) best supports a technician navigating the exploded view hands-on during a maintenance task?
5. How well does the approach generalise across different device types and component scales?

In this project we aim to...
- Build a working prototype that converts photos of a physical device into a navigable 3D exploded view with selectable components.
- Validate the workflow's speed and accuracy against a real maintenance scenario.
- Conduct a UX evaluation with target users (field technicians) to assess usability and trust in the guidance provided.