. 2024 Jul 5;14:15537. doi: 10.1038/s41598-024-66543-7

Table 2.

Details of implementation specifications.

Hardware/deep learning framework	Training hyper-parameters	Data augmentation	Time (ms)/img
$□$ Tensorflow: 2.13.0, Keras: 2.13.1, Cuda: 12.4, NVIDIA A100 40GB GPU $□$ Intel Core Silver 4316 CPU x86_64, 2.30 GHz 128 GM RAM	$□$ Img size: 224 $\times$ 224, Batch: 8 $□$ Optimizer: SGD, loss: categorical cross-entropy, learning rate: 0.007	$□$ Gaussian noise $□$ Random flip, rotation: 20, scale: 0.20, translation: 0.20	Using ResNet50 $□$ Train: 15.4 $□$ Inference: 5.8

Hardware/deep learning framework

Training hyper-parameters

Data augmentation

Time (ms)/img

$□$ Tensorflow: 2.13.0, Keras: 2.13.1, Cuda: 12.4, NVIDIA A100 40GB GPU

$□$ Intel Core Silver 4316 CPU x86_64, 2.30 GHz 128 GM RAM

$□$ Img size: 224 $\times$ 224, Batch: 8

$□$ Optimizer: SGD, loss: categorical cross-entropy, learning rate: 0.007

$□$ Gaussian noise

$□$ Random flip, rotation: 20, scale: 0.20, translation: 0.20

Using ResNet50

$□$ Train: 15.4

$□$ Inference: 5.8