Table 4 Real-time inference metrics of SignKeyNet across Platforms.

From: A multistream attention based neural network for visual speech recognition and sign language understanding

Metric

Desktop (RTX 3080)

Jetson Xavier NX

Inference Time (ms/frame)

17.3

23.2

Frames Per Second (FPS)

57.8

43.1

Peak Memory Usage (MB)

382

462

Model Size (MB)

46

46