Open umtbudak opened 7 years ago
.csv is a file format for Kaggle,maybe you don't need it. I think this is useful to you: https://github.com/jocicmarko/ultrasound-nerve-segmentation/issues/10 you can make test_mask.npy first,then use this code,produce mask image.
thanks @Kaido0 !
thanks @Kaido0 for your help. firtly ı run data.py and it prodused 4 .npy file and then i checked them (from .npy array to image) there is no problem in imgs_id_test.npy, imgs_test.npy, imgs_mask_train.npy and imgs_train.npy Note that ı used 500 training and 50 testing images for 20 iteration. after the running train.py, imgs_mask_test.npy was built. But The entire test mask images looks similar even the same. I applied the same codes without changing. How do i solve this problem?
@umtbudak Sorry,I don't know,I use my own data.But there is another simple method you can use to show mask.npy. Maybe you can see it clearly~
import numpy as np
import matplotlib.pyplot as plt
index=1
test_mask_array = np.load('imgs_mask_test.npy')
plt.imshow(test_mask_array[index][0], cmap='gray')
plt.show()
@Kaido0 thank you for your explanation. I have another problem about test mask size. ı tried 300x300 for im_row and im_col. but there was error below ValueError: "concat" mode can only merge layers with matching output shapes except for the concat axis. Layer shapes: [(None, 256, 36, 36), (None, 256, 37, 37)]
I did the process steps as you said. How can I perform testing? what does submission.csv file' mean? I could not create test mask.