WaymoBased on real interviews
Evaluate similar-scene retrieval
Measure whether a scene-search system helps researchers find useful examples for driving datasets.
Unlock with Full AccessThe challenge
Researchers search a large video archive with a query clip, then review returned clips for a training set. Adjacent windows can represent the same driving event, so a convincing result list may contain less useful evidence than it appears. Use the supplied audit setup to define a release scorecard for relevance and operating performance.
What you'll practise
Useful retrieval
Measure the relevance and distinct events researchers can actually add to a dataset.
Release criteria
Connect quality and performance measures to the decision to use the new model.
Retrieval model and audit setupCase materials included in the full challenge
Your scorecard
Define each measure and the decision it would support.
The complete brief and feedback are included with Full Access.