Table 2 Model complexity and computational cost.
From: Trustworthy pneumonia detection in chest X-ray imaging through attention-guided deep learning
Metric | Value | Comment |
|---|---|---|
Total parameters | 8,631,042 | Number of learnable weights in the entire neural network; indicates model size and expressive power |
Estimated MACs (THOP) | 4433.96 M | Multiply-Accumulate Operations (in millions) per forward pass, as measured by the THOP library |
FLOPs (ptflops) | 4.48 G | Floating-Point Operations per inference (in billions), estimated via ptflops for standardized comparison |
Parameters (ptflops) | 8.63 M | Model parameter count as independently estimated by ptflops; serves as a consistency check |
Approx GPU memory usage | 3760.00 MB | Approximate graphics memory required for model training/inference; critical for hardware deployment |
Avg training time per epoch (1 sample) | 0.0902 s | Time required to train on a single sample for one epoch—lower values indicate higher training speed |
Approx energy consumption (1 epoch, 1 sample) | 0.0038 Wh | Estimated electric energy consumed, useful for assessing energy efficiency and environmental impact |
