Model params 513 MB
Estimates for a single full pass of model at input size 384 x 384:
- Memory required for features: 426 MB
- Flops: 125 GFLOPs
Estimates are given below of the burden of computing the score_fr features in the network for different input sizes using a batch size of 128:
| input size | feature size | feature memory | flops |
|---|---|---|---|
| 192 x 192 | 7 x 7 x 21 | 22 GB | 7 TFLOPs |
| 384 x 384 | 13 x 13 x 21 | 49 GB | 16 TFLOPs |
| 576 x 576 | 19 x 19 x 21 | 87 GB | 29 TFLOPs |
| 768 x 768 | 25 x 25 x 21 | 136 GB | 46 TFLOPs |
| 960 x 960 | 31 x 31 x 21 | 196 GB | 68 TFLOPs |
| 1152 x 1152 | 37 x 37 x 21 | 267 GB | 93 TFLOPs |
A rough outline of where in the network memory is allocated to parameters and features and where the greatest computational cost lies is shown below. The x-axis does not show labels (it becomes hard to read for networks containing hundreds of layers) - it should be interpreted as depicting increasing depth from left to right. The goal is simply to give some idea of the overall profile of the model:
