Are you sure you want to delete this task? Once this task is deleted, it cannot be recovered.
langdu 574825ad03 | 2 years ago | |
---|---|---|
configs | 2 years ago | |
deploy | 2 years ago | |
images | 2 years ago | |
log | 2 years ago | |
paddleseg | 2 years ago | |
test_tipc | 2 years ago | |
tools | 2 years ago | |
LICENSE | 2 years ago | |
README.md | 2 years ago | |
export.py | 2 years ago | |
predict.py | 2 years ago | |
requirements.txt | 2 years ago | |
setup.py | 2 years ago | |
train.py | 2 years ago | |
val.py | 2 years ago |
本项目基于paddlepaddle框架复现了BiSeNet语义分割模型,BiSeNet引入Spatial Path和Context Path,Spatial Path中步长较小来得到高分辨率的特征图,Context Path引入快速的下采样机制来增加感受野,最后使用特征融合模块将Spatial Path和Context Path的特征图进行有效融合。。
论文:
项目参考:
在CityScapes val数据集的测试效果如下表。
NetWork | steps | opt | image_size | batch_size | dataset | memory | card | mIou | config | weight | log |
---|---|---|---|---|---|---|---|---|---|---|---|
BiSeNet | 160K | SGD | 1024x512 | 4 | CityScapes | 32G | 4 | 75.19 | bisenetv1_cityscapes_1024x512_160k.yml | link | log |
硬件: Tesla V100 * 4
框架:
# clone this repo
git clone https://github.com/justld/BisNetV1_paddle.git
cd BisNetV1_paddle
安装第三方库
pip install -r requirements.txt
单卡训练:
python train.py --config configs/bisenetv1/bisenetv1_cityscapes_1024x512_160k.yml --do_eval --use_vdl --log_iter 100 --save_interval 1000 --save_dir output
多卡训练:
python -m paddle.distributed.launch train.py --config configs/bisenetv1/bisenetv1_cityscapes_1024x512_160k.yml --do_eval --use_vdl --log_iter 100 --save_interval 1000 --save_dir output
output目录下包含已经训练好的模型参数以及对应的日志文件。
python val.py --config configs/bisenetv1/bisenetv1_cityscapes_1024x512_160k.yml --model_path
在linux下,进入BisNetV1_paddle文件夹,运行命令
bash test_tipc/prepare.sh ./test_tipc/configs/bisenetv1_small/train_infer_python.txt 'lite_train_lite_infer'
bash test_tipc/test_train_inference_python.sh ./test_tipc/configs/bisenetv1_small/train_infer_python.txt 'lite_train_lite_infer'
代码结构
├─configs
├─images
├─output
├─paddleseg
├─test_tipc
│ export.py
│ predict.py
│ README.md
│ README_CN.md
│ requirements.txt
│ setup.py
│ train.py
│ val.py
说明
1、本项目在Aistudio平台,使用Tesla V100 * 4 脚本任务训练16K miou达到75.19%。
2、本项目基于PaddleSeg开发。
相关信息:
信息 | 描述 |
---|---|
作者 | 郎督 |
日期 | 2021年11月 |
框架版本 | PaddlePaddle==2.2.0 |
应用场景 | 语义分割 |
硬件支持 | GPU、CPU |
在线体验 | notebook, Script |
使用Paddle复现BisNetV1
Python Java Shell Text Gradle other
Dear OpenI User
Thank you for your continuous support to the Openl Qizhi Community AI Collaboration Platform. In order to protect your usage rights and ensure network security, we updated the Openl Qizhi Community AI Collaboration Platform Usage Agreement in January 2024. The updated agreement specifies that users are prohibited from using intranet penetration tools. After you click "Agree and continue", you can continue to use our services. Thank you for your cooperation and understanding.
For more agreement content, please refer to the《Openl Qizhi Community AI Collaboration Platform Usage Agreement》