shartoo / luna16_multi_size_3dcnn

An implement of paper "Multi-level Contextual 3D CNNs for False Positive Reduction in Pulmonary Nodule Detection"
113 stars 45 forks source link

论文结果复现 #24

Open Itsanewday opened 2 years ago

Itsanewday commented 2 years ago

up主您好,请问之前您复现过该论文的结果吗,运行速度如何?烦请回复,谢谢

shartoo commented 2 years ago

之前复现过,但是代码可能不是这个,这个代码好像生成样本时会出错。这个代码只是个中间尝试,效果一般,推荐改用C3D网络

Itsanewday commented 2 years ago

麻烦问下你说的C3D是指什么,有共享的链接地址吗?谢谢

shartoo commented 2 years ago

介绍: https://sh-tsang.medium.com/paper-c3d-learning-spatiotemporal-features-with-3d-convolutional-networks-video-classification-72b49adb4081

论文: https://arxiv.org/abs/1412.0767

Itsanewday commented 2 years ago

此前一直陷入困境,非常感谢您的帮助,祝好!

lander1003 commented 2 years ago

介绍: https://sh-tsang.medium.com/paper-c3d-learning-spatiotemporal-features-with-3d-convolutional-networks-video-classification-72b49adb4081

论文: https://arxiv.org/abs/1412.0767

你好,想问下C3D用来做结节检测准确率高吗

shartoo commented 2 years ago

75%左右吧,在luna16的数据集基础之上,具体的优化可能需要自己去整合数据和调参了。C3D只是个base model,抽取特征用,后续可以自己拼接全连接层或者自己添加新的网络层。