Ready for the primary review
Higher-confidence windows supported by HUD combat evidence or corroborated detector signals.
Fight detection
Find high-confidence Apex fights, then check the footage.
ApexLens separates accepted fight windows from lower-confidence candidates and records the exact interval so a player or coach can check what belongs in the review.
- Precision-first accepted tier
- Candidate tier stays visible
- Exact source intervals
Two review tiers
A candidate is useful without pretending it is certain.
Accepted windows cross a higher confidence threshold. Candidate windows preserve weaker signals for inspection. Keeping the two separate lets the interface prefer clean accepted results without hiding every uncertain moment.
Worth a human look
Lower-confidence local signals remain available because movement, looting, and visual noise can resemble combat.
Expanded evaluation snapshot0.963accepted-window precision
The tradeoff0.361accepted-window recall
Usually clean does not mean complete.
On the expanded labeled set, accepted combined windows produced 26 true positives, one false positive, and 46 false negatives. That is why ApexLens can describe high-confidence windows, but cannot promise to find every fight.
Accepted
23:53–24:01 · 0.705
Candidates
07:58–08:24 · 0.695