Table 6.

Diagnosis performance of the proposed method at a scaling rate of 0.01 and consumed system resources compared with existing methods.

0.01 × MobileNet-v2	MACs	Parameters	Average Accuracy
Original	602,096	46,056	99.58%
Sparsity rate: 0.1	596,816	41,304	99.58%
Sparsity rate: 0.2	592,776	37,668	99.38%
Sparsity rate: 0.3	589,174	34,788	98.95%
LeNet5-based methods	MACs (LeNet5)	Parameters (LeNet5)	Average Accuracy
Method [21]	668,272	72,376	98.74%
Method [20]	668,272	72,376	94.20%