Table 2 Without loading the pre-training model, our model mIoU is slightly higher than ICNet
From: AtICNet: semantic segmentation with atrous spatial pyramid pooling in image cascade network
| Â | ICNet | AtICNet |
|---|---|---|
mIoU (%) | 24.9 | 25.1 |
Training memory (M) | 4553 | 8649 |
Evaluation memory (M) | 7111 | 7655 |
Evaluation time (S) | 127 | 136 |