SG-VLM

Reported on 4 benchmarks across 2 tasks

Note: results are matched by exact model name. Different papers may use the same name for different model variants.

Robots2 results

Activity RecognitiononMOMA-LRG
Activity Classification Accuracy (5-shot 5-way)· uses extra data
92.5
best: 97.9 (Name Tuning)
Activity RecognitiononMOMA-LRG
Subactivity Classification Accuracy (5-shot 5-way)· uses extra data
32.7
best: 78.2 (Name Tuning)

Time Series2 results

Action RecognitiononMOMA-LRG
Activity Classification Accuracy (5-shot 5-way)· uses extra data
92.5
best: 97.9 (Name Tuning)
Action RecognitiononMOMA-LRG
Subactivity Classification Accuracy (5-shot 5-way)· uses extra data
32.7
best: 78.2 (Name Tuning)