Are you sure you want to delete this task? Once this task is deleted, it cannot be recovered.
ych 3943e7814b | 1 year ago | |
---|---|---|
Base | 1 year ago | |
DataLoader | 1 year ago | |
Model/Deeplabv3_plus | 1 year ago | |
Utils | 1 year ago | |
configs | 1 year ago | |
README.md | 1 year ago | |
main.py | 1 year ago | |
train.py | 1 year ago |
[CVPR'22][Perturbed and Strict Mean Teachers for Semi-supervised Semantic Segmentation]
by Yuyuan Liu, Yu Tian, Yuanhong Chen, Fengbei Liu, Vasileios Belagiannis and Gustavo Carneiro
Computer Vision and Pattern Recognition Conference (CVPR), 2022
原作者是由pytorch构建,github地址
1.数据集使用Pascal VOC12 dataset,原始数据下载地址:origion dataset,SegmentationClassAug
2.预训练模型:[Pre-trained models](链接:https://pan.baidu.com/s/1xzY--3prSYDdgjwwSKc5ZA?pwd=563o 提取码:563o)
3.训练参数设置:在数据集分类使用Pascal voc 2012数据划分1/2(732) , Deeplabv3+(resnet 101)
单卡: batch_size=16 (8 labelled, 8 unlabelled data) , learning rate 2.5e-3,epoch 80,unsup_weight 0.6,warm_up 5,semi_p_th 0.6;
训练细节:VOC12 dataset
复现精度:miou 78.05%
训练模型:[model](链接:https://pan.baidu.com/s/1UK0awndzzbKTLh415oGMNg?pwd=myqo 提取码:myqo)
1.将数据集放在VocCode目录下,修改路径:DataLoader/voc.py
2.预训练模型放在Model/Deeplabv3_plus/Backbones/pretrained下,修改路径:Model/Deeplabv3_plus/EntireModel.py
3.运行训练脚本:python main.py
目前单卡已经复现完成,多卡还不完善
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》