VGG type Network
ConvBlock (32, 360, 360)
ConvBlock (64, 180, 180)
ConvBlock (128, 90, 90)
ConvBlock (256, 45, 45)
ConvBlock (512, 22, 22)
Global Average Pooling
Dense(2)
ConvBlock
Conv3x3 + BN + ELU
Conv3x3 + BN + ELU
MaxPool (2, 2)
Batch: 32
Optimizer: Adam
LR: 1e-4
5 fold
Loss: L1
Augmentation: flips
Training
A
B
C
D
E
F
BCE = - \sum_i \left( y_i \ln (p_i) + (1 - y_i) \ln(1 - p_i)\right)
B
C
E
=
−
∑
i
(
y
i
ln
(
p
i
)
+
(
1
−
y
i
)
ln
(
1
−
p
i
)
)
BCE = - \sum_i \left( y_i \ln (p_i) + (1 - y_i) \ln(1 - p_i)\right)
B
C
E
=
−
∑
i
(
y
i
ln
(
p
i
)
+
(
1
−
y
i
)
ln
(
1
−
p
i
)
)
DICE = 2 \frac {\sum_i y_i p_i} {\sum_u y_i + \sum p_i}
D
I
C
E
=
2
∑
i
y
i
p
i
∑
u
y
i
+
∑
p
i
DICE = 2 \frac {\sum_i y_i p_i} {\sum_u y_i + \sum p_i}
D
I
C
E
=
2
∑
u
y
i
+
∑
p
i
∑
i
y
i
p
i
LOSS = BCE - \ln \left( DICE \right)
L
O
S
S
=
B
C
E
−
ln
(
D
I
C
E
)
LOSS = BCE - \ln \left( DICE \right)
L
O
S
S
=
B
C
E
−
ln
(
D
I
C
E
)
BCE = - \sum_i \left( y_i \ln (\hat{y_i}) + (1 - y_i) \ln(1 - \hat{y_i})\right)
B
C
E
=
−
∑
i
(
y
i
ln
(
y
i
^
)
+
(
1
−
y
i
)
ln
(
1
−
y
i
^
)
)
BCE = - \sum_i \left( y_i \ln (\hat{y_i}) + (1 - y_i) \ln(1 - \hat{y_i})\right)
B
C
E
=
−
∑
i
(
y
i
ln
(
y
i
^
)
+
(
1
−
y
i
)
ln
(
1
−
y
i
^
)
)
SoftJaccard = \frac{1}{n}\sum\limits_{i=1}^n\left(\frac{y_i\hat{y}_i}{y_{i}+\hat{y}_i-y_i\hat{y}_i}\right)
S
o
f
t
J
a
c
c
a
r
d
=
1
n
∑
i
=
1
n
(
y
i
y
^
i
y
i
+
y
^
i
−
y
i
y
^
i
)
SoftJaccard = \frac{1}{n}\sum\limits_{i=1}^n\left(\frac{y_i\hat{y}_i}{y_{i}+\hat{y}_i-y_i\hat{y}_i}\right)
S
o
f
t
J
a
c
c
a
r
d
=
n
1
i
=
1
∑
n
(
y
i
+
y
^
i
−
y
i
y
^
i
y
i
y
^
i
)
LOSS = \alpha \times BCE + (1 - \alpha) \times SoftJaccard
L
O
S
S
=
α
×
B
C
E
+
(
1
−
α
)
×
S
o
f
t
J
a
c
c
a
r
d
LOSS = \alpha \times BCE + (1 - \alpha) \times SoftJaccard
L
O
S
S
=
α
×
B
C
E
+
(
1
−
α
)
×
S
o
f
t
J
a
c
c
a
r
d
CCE = - \frac {1} {n} \sum_{c=1}^7 \sum_{i=1}^{n} y_i^c \ln (\hat{y_i}^c)
C
C
E
=
−
1
n
∑
c
=
1
7
∑
i
=
1
n
y
i
c
ln
(
y
i
^
c
)
CCE = - \frac {1} {n} \sum_{c=1}^7 \sum_{i=1}^{n} y_i^c \ln (\hat{y_i}^c)
C
C
E
=
−
n
1
∑
c
=
1
7
∑
i
=
1
n
y
i
c
ln
(
y
i
^
c
)
SoftJaccard =\frac{1}{n}\sum_{c=1}^7w_c\sum\limits_{i=1}^n\left(\frac{y_i^c\hat{y}^c_i}{y_{i}^c+\hat{y}^c_i-y_i^c\hat{y}_i^c}\right)
S
o
f
t
J
a
c
c
a
r
d
=
1
n
∑
c
=
1
7
w
c
∑
i
=
1
n
(
y
i
c
y
^
i
c
y
i
c
+
y
^
i
c
−
y
i
c
y
^
i
c
)
SoftJaccard =\frac{1}{n}\sum_{c=1}^7w_c\sum\limits_{i=1}^n\left(\frac{y_i^c\hat{y}^c_i}{y_{i}^c+\hat{y}^c_i-y_i^c\hat{y}_i^c}\right)
S
o
f
t
J
a
c
c
a
r
d
=
n
1
∑
c
=
1
7
w
c
i
=
1
∑
n
(
y
i
c
+
y
^
i
c
−
y
i
c
y
^
i
c
y
i
c
y
^
i
c
)
LOSS = \alpha \times CCE + (1 - \alpha) \times SoftJaccard
L
O
S
S
=
α
×
C
C
E
+
(
1
−
α
)
×
S
o
f
t
J
a
c
c
a
r
d
LOSS = \alpha \times CCE + (1 - \alpha) \times SoftJaccard
L
O
S
S
=
α
×
C
C
E
+
(
1
−
α
)
×
S
o
f
t
J
a
c
c
a
r
d
Jaccard
Public Test
0.636
Private Test
0.613
Jaccard
Public Test
0.493
Private Test
0.523
Made with Slides.com