ResearchBenchmarksAction Recognition on MammalNet (test)Action Recognition on MammalNet (test)Follow56.6Top-1 AccuracyUAS46.248.951.654.3Feb 10, 2026Evaluation resultsMethodMethodLinksTop-1 AccuracyMCAUASBackbone=VST [23], Pre...Backbone=VST [23], Pre-trained=K-400, Training Strategy=Linear Probe, Training Time (hr)=8.3, #Params (K)=12.32026.0256.643.2MammalNetBackbone=MViTv2 [21],...Backbone=MViTv2 [21], Pre-trained=K-400, Training Strategy=Full Fine-tune, Training Time (hr)=248.8, #Params (K)=51,028.72026.0246.637.8