The naming convention for models uses the following format:
COVID-Net Assistant <seed design>
For example, a COVID-Net Assistant model generated with Res-CNN-S seed design (i.e. small model based on residual blocks), which was trained on the dataset's verified-only split would be called "COVID-Net Assistant Res-CNN-S".
These models are trained and tested on the Verified-Only Split of Covid19-Cough dataset
Seed Design used to generate final COVID-Net Assistant architectures via Generative Synthesis:
| Seed Design | AUC | Latency WORKSTN (ms) | Latency SBC (ms) | Parameters (K) | FLOPs (M) | AUC |
|---|---|---|---|---|---|---|
| CNN | 0.7925 / 0.9381 | 3.84 | 32.8 | 65.3 | 491.3 | 0.7815 / 0.9508 |
| Res-CNN-S | 0.7912 / 0.9285 | 3.06 | 23.5 | 367.8 | 249.3 | 0.7828 / 0.9349 |
| Res-CNN-M | 0.7897 / 0.9251 | 4.19 | 29.6 | 927.5 | 367.9 | 0.7887 / 0.9394 |
| Res-CNN-L | 0.7942 / 0.9204 | 5.2 | 48.5 | 1158.2 | 675.6 | 0.8039 / 0.9388 |
| DW-CNN-S | 0.7786 / 0.9279 | 3.47 | 8.84 | 18 | 22.3 | 0.786 / 0.933 |
| DW-CNN-M | 0.7722 / 0.941 | 4.19 | 10.27 | 142.1 | 27.5 | 0.7832 / 0.9304 |
| DW-CNN-L | 0.7829 / 0.9322 | 20.6 | 52.5 | 159 | 149.8 | 0.7692 / 0.9305 |
| Model Name | AUC | Latency WORKSTN (ms) | Latency SBC (ms) | Parameters(K) | FLOPs(M) | Pre-trained Weights |
|---|---|---|---|---|---|---|
| COVID-Net Assistant CNN | 0.7815 / 0.9508 | 2.4 / 1.72 | 10.7 / 9.44 | 6.6 / 4.6 | 63.3 / 45.8 | All-Data / Verified-Only |
| COVID-Net Assistant Res-CNN-S | 0.7828 / 0.9349 | 2.29 / 2.22 | 8.76 / 10.6 | 74.3 / 76.3 | 25.2 / 59.7 | All-Data / Verified-Only |
| COVID-Net Assistant Res-CNN-M | 0.7887 / 0.9394 | 2.64 / 2.61 | 11.2 / 12.1 | 161.4 / 228.8 | 39.7 / 55.6 | All-Data / Verified-Only |
| COVID-Net Assistant Res-CNN-L | 0.8039 / 0.9388 | 3.24 / 3.05 | 21.8 / 13.6 | 254.8 / 229.0 | 175.9 / 66.7 | All-Data / Verified-Only |
| COVID-Net Assistant DW-CNN-S | 0.786 / 0.933 | 2.34 / 2.31 | 4.94 / 5.39 | 3.5 / 3.9 | 3.9 / 4.5 | All-Data / Verified-Only |
| COVID-Net Assistant DW-CNN-M | 0.7832 / 0.9304 | 3.06 / 3.18 | 6.74 / 6.98 | 25.2 / 23.7 | 6.2 / 6.5 | All-Data / Verified-Only |
| COVID-Net Assistant DW-CNN-L | 0.7692 / 0.9305 | 14 / 14.1 | 35.1 / 35.2 | 34.1 / 35.2 | 71 / 71.1 | All-Data / Verified-Only |
Format follows: "evaluiation on All-Data Split / evaluiation on Verified-Only Split"
