AI System for Classifying Site Photos by Work Item and Area

The problem

Photographs may be filed without labels during day-to-day operations. When photographs, capture times, locations, area lists and work schedules are scattered across different screens, documents or team records, issues do not become visible in time.

This fragmented data delays preparation of filing suggestions and requires staff to search for supporting records again by hand. Before making a decision, the site engineer has to complete missing fields and verify conflicts one by one.

A common exception is that similar site areas may not be distinguishable with certainty. If an incomplete or incorrect workflow runs without checking this context, a photograph may be treated as evidence for the wrong work item; the output therefore cannot be applied directly.

What we set out to improve

• Rate of corrected site-area or work-item labels
• Rate of outputs corrected or rejected by the site engineer
• Time between the source record and the review output

How the system works

1. 01 Define the review scope and the authorised site engineer role
2. 02 Receive the inputs: photograph, capture time, location, area list and work schedule
3. 03 Check source identity, date, version and data freshness
4. 04 Extract metadata and visible indicators
5. 05 Generate candidates using the site area and work schedule
6. 06 Check the exception: similar site areas may not be distinguishable with certainty
7. 07 Place low-confidence or conflicting records in a separate human review queue
8. 08 Write the source identity, rule, output and time to the audit trail
9. 09 Have the site engineer approve, correct, defer or reject the suggestion

Methods we used

• Pre-classification of images and metadata matching
• Source, format, version and required-field checks for the photograph, capture time, location, area list and work schedule
• A missing-data warning instead of a prediction when inputs are absent or conflicting
• A separate check for the exception outside the main rule
• Joint storage of the source identity, rule result, filing suggestion and human decision

Where people stay involved

Data and security

Access to photographs, capture times, locations, area lists and work schedules is restricted to the minimum permissions required for each role. Identifiers and sensitive fields are masked where possible, and only the necessary section of a file is sent to the model rather than the complete raw file. Access, output and the site engineer's decision are logged; production use does not begin until the organisation has approved the retention period and data location.

Who this suits

Frequently asked questions