Collected sources and patterns will appear here. Add from search, explore, or the patterns library.
List<Logits> -> AveragedLogits
Aggregate predicted classification logits from multiple spatial crops and temporal clips of a single video via averaging.
Problem it solves
Single-view inference ignores spatial-temporal boundaries, lowering validation accuracy and model stability.
Consumes
Emits
The real projects this mechanism was found in. Attribution is the point — this is how the best teams actually do it.