search

juliandewit / kaggle_ndsb2017

Kaggle datascience bowl 2017
MIT License
622 stars 290 forks source link

some questions about the evaluation of the malignancy of the nodules #66

Closed fulingly closed 4 years ago

fulingly commented 4 years ago

The second task of the 3d cnn network is to evaluate the malignancy of the nodules, out_malignancy = Convolution3D(1, 1, 1, 1, activation=None, name="out_malignancy_last")(last64) out_malignancy = Flatten(name="out_malignancy")(out_malignancy) but in the step3_predict_nodules.py,the diameter_mm = round(p[1][i][0], 4),why the second output of the network is the diameter of the nodule,not the malignancy,if I understanding something wrong.I want your help desirely,thank you very much.

juliandewit commented 4 years ago

I started with diameter as proxy for cancer. Later I found malignancy labels. I think this was a naming error