I've a data set of the following format. It's a binary classification problem (0,1). For this I've made a custom generator down below but couldn't understand how should I properly encode the label there.
- images/
- a.png
- b.png
- c.png
- images.csv
img_Id
-
-
label
a.png
-
-
1
b.png
-
-
0
c.png
-
-
1
Generator
class generate(Sequence):
def __init__(self, data, dim, shuffle, batch_size):
self.data = data
.....
def __len__(self):
return int(np.floor(len(self.data)/self.batch_size))
def __getitem__(self, index):
batch_idx = self.indices[index*self.batch_size:(index+1)*self.batch_size]
idx = [self.list_idx[k] for k in batch_idx]
Data = np.empty((self.batch_size, *self.dim, 1))
Target = np.empty((self.batch_size, 1), dtype = int)
for i, k in enumerate(idx):
# load the image file using cv2
image = cv2.imread(self.data['image_name'][k])
image = cv2.resize(image, self.dim)
# expand the axises
Data[i,:, :, :] = image
Target[i,:] = ??????????????????????????????
return Data Target
def on_epoch_end(self):
-----
The model should output with sigmoid activation on the last layer. How properly I can encode label here and match properly with the image file?? Is there any more convenient way to do this??
Additional Query:
Why sometimes predict_generator doesn't return all samples during operation. Asked here but I din't get the answer. :(
I am trying to take this issue as a learning opportunity. So just one more related question. If My images.label attributes contains multi-classes, what should I changes to make it work. Or, multi-label?
I've a data set of the following format. It's a binary classification problem (0,1). For this I've made a custom generator down below but couldn't understand how should I properly encode the label there.
Generator
Call
The model should output with
sigmoid
activation on the last layer. How properly I can encode label here and match properly with the image file?? Is there any more convenient way to do this??Additional Query:
Why sometimes
predict_generator
doesn't return all samples during operation. Asked here but I din't get the answer. :(I am trying to take this issue as a learning opportunity. So just one more related question. If My
images.label
attributes contains multi-classes, what should I changes to make it work. Or, multi-label?Thanks in advance.