|
| 1 | +# Metadata |
| 2 | +author: Y-T-G |
| 3 | +task: pose |
| 4 | +keywords: [transformer, swin] |
| 5 | +description: Swin-T pretrained backbone with Pose head. |
| 6 | +flops: 67.2 |
| 7 | +parameters: 32432152 |
| 8 | +min_version: 8.3.64 |
| 9 | + |
| 10 | +# Model |
| 11 | +nc: 1 |
| 12 | +strides: [8, 16, 32] |
| 13 | +kpt_shape: [17, 3] |
| 14 | +backbone: |
| 15 | + # [from, repeats, module, args] |
| 16 | + - [-1, 1, TorchVision, [768, "swin_t", "DEFAULT", True, 5, True]] |
| 17 | + - [0, 1, Index, [192, 4]] # (1, 80, 80, 192) - 1 |
| 18 | + - [-1, 1, torchvision.ops.Permute, [[0, 3, 1, 2]]] # (1, 192, 80, 80) - 2 |
| 19 | + - [0, 1, Index, [384, 6]] # (1, 40, 40, 384) - 3 |
| 20 | + - [-1, 1, torchvision.ops.Permute, [[0, 3, 1, 2]]] # (1, 384, 40, 40) - 4 |
| 21 | + - [0, 1, Index, [768, 8]] # (1, 20, 20, 768) - 5 |
| 22 | + - [-1, 1, torchvision.ops.Permute, [[0, 3, 1, 2]]] # (1, 768, 20, 20) - 6 |
| 23 | +head: |
| 24 | + - [[2, 4, 6], 1, Pose, [nc, kpt_shape]] |
0 commit comments