Video Understanding on Video Understanding Benchmark
58AccuracyLiteFrame
Evaluation Results
| Method | Links | |
|---|---|---|
| LiteFrameVision Params=87M, Tok./Frame=16, Frames=32, Total Tok.=512, Vis. (ms)=30.1, LLM (ms)=61.5, Total (ms)=91.92026.05 | 58 | |
| VideoPandaVision Params=45M, Tok./Frame=272, Frames=32, Total Tok.=8704, Vis. (ms)=36.5, LLM (ms)=345.7, Total (ms)=382.42026.05 | 49.2 | |
| FastVLMVision Params=125M, Tok./Frame=49, Frames=32, Total Tok.=1568, Vis. (ms)=98.3, LLM (ms)=132.9, Total (ms)=231.52026.05 | 47.6 |