fazil
/
pytorch-dann-resnet
1
0
Fork 0
Code Issues Pull Requests Projects Releases Wiki Activity
A PyTorch implementation for paper Unsupervised Domain Adaptation by Backpropagation
30 Commits
1 Branch
0 Tags
152 KiB
 
Tree: 5e65946d13
wogong 5e65946d13
add bn layer, source only get 60%, while DA get unstable results, maximum 61%, very similar to source only. 		6 years ago
core add l2 norm. 6 years ago
datasets add bn layer, source only get 60%, while DA get unstable results, maximum 61%, very similar to source only. 6 years ago
experiments add bn layer, source only get 60%, while DA get unstable results, maximum 61%, very similar to source only. 6 years ago
models add bn layer, source only get 60%, while DA get unstable results, maximum 61%, very similar to source only. 6 years ago
utils add synsigns_gtsrb experiment, update core codes 6 years ago
.gitignore minor update, remove some params 7 years ago
LICENSE Initial commit 7 years ago
README.md minor update 6 years ago

README.md

PyTorch-DANN

A PyTorch implementation for paper Unsupervised Domain Adaptation by Backpropagation

InProceedings (icml2015-ganin15)
Ganin, Y. & Lempitsky, V.
Unsupervised Domain Adaptation by Backpropagation
Proceedings of the 32nd International Conference on Machine Learning, 2015

Environment

  • Python 3.6
  • PyTorch 1.0

Note

  • Config() 为针对特定任务的配置参数
  • MNISTmodel() 完全按照论文中的结构，但是 feature 部分添加了 Dropout2d()，实验发现是否添加 Dropout2d() 对于最后的性能影响很大。最后实验重现结果高于论文，因为使用了额外的技巧，这里还有值得探究的地方。
  • SVHNmodel() 无法理解论文中提出的结构，为自定义结构。最后实验重现结果完美。
  • MNIST-MNISTM: python mnist_mnistm.py
  • SVHN-MNIST: python svhn_mnist.py
  • Amazon-Webcam: python office.py 由于预训练网络的问题，无法复现

Result

MNIST-MNISTM SVHN-MNIST Amazon-Webcam Amazon-Webcam10
Source Only 0.5225 0.5490 0.6420 0.
DANN(paper) 0.7666 0.7385 0.7300 0.
This Repo Source Only - - - 0.
This Repo 0.8400 0.7339 0.6528 0.

Credit