Table 5.
performance comparison of the variants of VSL models on MedVidQA dataset.
| Models | IoU = 0.3 | IoU = 0.5 | IoU = 0.7 | mIoU | |
|---|---|---|---|---|---|
| Random Mode | 8.38 | 1.93 | 1.21 | 6.89 | |
| Random Guess | 7.74 | 3.22 | 0.64 | 5.96 | |
| VSL-BASE | FPL 400 | 19.35 | 6.45 | 3.22 | 18.08 |
| FPL 600 | 19.35 | 10.96 | 4.51 | 19.20 | |
| FPL 800 | 21.93 | 12.25 | 5.80 | 20.15 | |
| FPL 1000 | 21.93 | 7.74 | 3.87 | 18.86 | |
| FPL 1200 | 22.58 | 9.67 | 5.16 | 19.97 | |
| FPL 1400 | 25.16 | 8.38 | 4.51 | 19.33 | |
| VSL-QGH | 25.81 | 14.20 | 6.45 | 20.12 | |
Here FPL refers to the frame position length considered during training the respective models.