Table 6 Computational efficiency and training speed improvement of CondenseBlock compared to DenseBlock in one training epoch.

From: ID-insensitive deepfake detection model based on multi-attention mechanism

Metric

DensenBlock

CondenseBlock

Improvement(%)

Inference Time(s)

2252.10

2065.46

8.28

Memory

Usage(Mb)

11,081.84

9724.64

12.24

Back to article page