Understand the image in the context of the journey
A sharp photo is not necessarily a strong holiday image. Assessment therefore also considers composition, light, atmosphere and whether the image conveys a special moment.
Post text, capture data, location and itinerary complement the visual impression. This lets the AI identify contradictions, for example when the location or time does not match the described trip.
A reasoned assessment for editors
Editors receive a quality score and an assessment of the photo's strengths and weaknesses. The AI also flags uncertainty and cases where it recommends human review.
Editors can confirm or correct the assessment. The solution supports selection with a consistent foundation and highlights where closer review is useful.
Our own criteria, our own model operations
What makes a good community photo depends on its purpose. For the AIDA Lounge, we focus on authentic holiday moments and editorial image quality. These criteria can be adjusted and developed deliberately.
Running on our own hardware gives us control over the models and processing of assessment requests. This includes monitoring instances, handling failures and developing the system further.
Integrated into ongoing operations
Image assessment is part of the AIDA Lounge: its results feed into quality filters and post weighting. A review interface allows people to inspect and correct assessments.
The project shows how we integrate AI into existing applications, with domain-specific criteria, traceable results and opportunities for human intervention.