Table 1.
Layers of SunnyNet.
| Layer type | Kernel shape | Output shape | Learnable parameters |
|---|---|---|---|
| 3D Conv | (3, 3, 3) | ( − 1, 32, 400, 1, 1) | 5216 |
| 1D Conv | (3) | ( − 1, 32, 398) | 3104 |
| 1D MaxPool | (2) | ( − 1, 32, 199) | 0 |
| 1D Conv | (3) | ( − 1, 64, 197) | 6208 |
| 1D MaxPool | (2) | ( − 1, 64, 98) | 0 |
| 1D Conv | (3) | ( − 1, 128, 96) | 24 704 |
| 1D MaxPool | (2) | ( − 1, 128, 48) | 0 |
| Linear | – | ( − 1, 4, 700) | 28 881 500 |
| Dropout | – | ( − 1, 4, 700) | 0 |
| Linear | – | ( − 1, 2, 400) | 11 282 400 |
Notes. Input shape: ( − 1, 6, 400, 1, 1). Total parameters: 40 203 132. Parameter size: 153.36 MB. The layers are shown in order, from the first 3D convolutional layer to the output linear layer. The stride for 3D Conv, 1D Conv, and 1D MaxPool are one, one, and two respectively. The 3D Conv layer is also 0-padded in the z direction. The −1 dimension in the “Output Shape” column is a placeholder for however many input samples are in each training batch.
Current usage metrics show cumulative count of Article Views (full-text article views including HTML views, PDF and ePub downloads, according to the available data) and Abstracts Views on Vision4Press platform.
Data correspond to usage on the plateform after 2015. The current usage metrics is available 48-96 hours after online publication and is updated daily on week days.
Initial download of the metrics may take a while.