BiSeNetV1 (ECCV'2018)
@inproceedings{yu2018bisenet,
title={Bisenet: Bilateral segmentation network for real-time semantic segmentation},
author={Yu, Changqian and Wang, Jingbo and Peng, Chao and Gao, Changxin and Yu, Gang and Sang, Nong},
booktitle={Proceedings of the European conference on computer vision (ECCV)},
pages={325--341},
year={2018}
}
Segmentor | Pretrain | Backbone | Crop Size | Schedule | Train/Eval Set | mIoU | Download |
---|---|---|---|---|---|---|---|
FCN | ImageNet-1k-224x224 | BiSeNetV1, R-18-D32 | 1024x1024 | LR/POLICY/BS/EPOCH: 0.05/poly/16/860 | train/val | 75.76% | cfg | model | log |
FCN | ImageNet-1k-224x224 | BiSeNetV1, R-50-D32 | 1024x1024 | LR/POLICY/BS/EPOCH: 0.05/poly/16/860 | train/val | 77.78% | cfg | model | log |
You can also download the model weights from following sources:
- BaiduNetdisk: https://pan.baidu.com/s/1gD-NJJWOtaHCtB0qHE79rA with access code s757