search

wentaozhu / deep-mil-for-whole-mammogram-classification

Zhu, Wentao, Qi Lou, Yeeleng Scott Vang, and Xiaohui Xie. "Deep Multi-instance Networks with Sparse Label Assignment for Whole Mammogram Classification." MICCAI 2017.
MIT License
114 stars 37 forks source link

what the “mil”, "mymil","mysparsemil","mysoftmil" stand for? Thanks! #10

Closed guyucowboy closed 6 years ago

guyucowboy commented 6 years ago

Hi, wentao zhu Great work! Thanks for sharing you codes! what the “mil”, "mymil","mysparsemil","mysoftmil" stand for? mil = Multiple Instance Learning How about others? Best regards. Thanks! Gu Yu

wentaozhu commented 6 years ago

MIL is with max-pooling inference mymil is with label assignment inference mysparsemil is with sparisity inference mysoftmil is todo inference, I think, it is sparity with label assignment. did not appear in the paper.

You can find the details in the MICCAI17 paper, Deep multi-instance networks with sparse label assignment for whole mammogram classification https://arxiv.org/pdf/1612.05968.pdf

After that, you can check the code carefully about it.

guyucowboy commented 6 years ago

Hi, wentao zhu Thanks for your reply! I have an another question to ask you. Why do you use alexnet rather than vgg-16, vgg-19 in convnets-keras package? Thanks. Best regards. Gu Yu

wentaozhu commented 6 years ago

Dear Gu,

You can definitely try the VGG-16 or VGG-19. At that, I did not have powerful machine and the dataset is not big.

Wentao

guyucowboy commented 6 years ago

Hi, wentao zhu. Thanks foy your reply! Gu Yu