Table 2.
Comparison of the baseline and proposed models in terms of computational efficiency, memory, and time*.
| Model | Params | FLOPs | Size | Training | Inference |
|---|---|---|---|---|---|
| Name | (million) | (billion) | (mb) | (mins)** | (secs) |
| Res2Net | 23.53 | 1.75 | 90.21 | 2.28 | 56.80 |
| LDR | 21.72 | 1.40 | 92.22 | 1.98 | 42.21 |
| ESM | 23.00 | 1.67 | 77.34 | 2.19 | 59.43 |
| VIN | 27.16 | 2.1 | 95.53 | 2.35 | 70.10 |
| ShuffleNetV2 | 1.23 | 1.24 | 5.10 | 1.04 | 34.53 |
| MnasNet | 3.12 | 0.93 | 12.41 | 1.31 | 26.59 |
| MobileNetV3 | 2.24 | 0.73 | 8.96 | 0.53 | 8.75 |
| Ours | 0.86 | 0.65 | 3.44 | 0.47 | 7.53 |
Number of trainable parameters, training, and inference time may differ depending on the datasets characteristics. *This information is based on experiments using 32 GB NVIDIA Tesla V100-SXM2 GPU. **Average training time per epoch.
Significant values are in [bold].