Skip to content

Commit 9a1fc10

Browse files
Cooper1973Cooper1973
committed
AmygdalaGo-BOLT
1 parent 8766722 commit 9a1fc10

File tree

174 files changed

+20282
-2
lines changed

Some content is hidden

Large Commits have some content hidden by default. Use the searchbox below for content that may be hidden.

174 files changed

+20282
-2
lines changed

Diff for: ‎H3/AmygdalaGo-BOLT/.ipynb_checkpoints/cmd-checkpoint.sh

+1-1
Original file line numberDiff line numberDiff line change
@@ -25,4 +25,4 @@ python test.py --model utnetv2 --dimension 3d --dataset acdc --batch_size 1 --un
2525
python inference.py --model unet --dimension 3d --dataset acdc --batch_size 1 --unique_name acdc_3d_unet --gpu 0
2626

2727

28-
python inference.py --model utnetv2 --dimension 3d --dataset acdc --batch_size 1 --unique_name acdc_3d_utnetv2 --gpu 0
28+
python inference.py --model utnetv2 --dimension 3d --dataset acdc --batch_size 1 --unique_name acdc_3d_utnetv2 --gpu 0

Diff for: ‎H3/AmygdalaGo-BOLT/app_demo_data/input/001_1.nii.gz renamed to ‎H3/AmygdalaGo-BOLT/app_demo_data/inputs/001_1.nii.gz

File renamed without changes.

Diff for: ‎H3/AmygdalaGo-BOLT/checkpoint_v1.0/bolt.pth

80.8 MB
Binary file not shown.

Diff for: ‎H3/AmygdalaGo-BOLT/config/acdc/attention_unet_2d.yaml

+44
Original file line numberDiff line numberDiff line change
@@ -0,0 +1,44 @@
1+
#DATA
2+
data_root: dataset/acdc_2d
3+
classes: 4
4+
modality: mri
5+
6+
7+
#MODEL
8+
arch: attention_unet
9+
in_chan: 1
10+
base_chan: 32
11+
12+
13+
#TRAIN
14+
epochs: 150
15+
training_size: [256, 256] # training crop size
16+
start_epoch: 0
17+
18+
seed: 0
19+
k_fold: 5
20+
21+
optimizer: adamw
22+
base_lr: 0.0005
23+
betas: [0.9, 0.999]
24+
weight_decay: 0.05 # weight decay of SGD optimizer
25+
weight: [0.5, 1, 1, 1] # weitght of each class in the loss function
26+
rlt: 1 # relation between CE and Dice loss
27+
28+
scale: 0.3 # scale for data augmentation
29+
rotate: 180 # rotation angle for data augmentation
30+
translate: 0
31+
gaussian_noise_std: 0.02
32+
additive_brightness_std: 0.7
33+
gamma_range: [0.5, 1.6]
34+
35+
36+
#VALIDATION
37+
ema: True
38+
ema_alpha: 0.99
39+
val_frequency: 10
40+
41+
42+
43+
#INFERENCE
44+
sliding_window: False

Diff for: ‎H3/AmygdalaGo-BOLT/config/acdc/attention_unet_3d.yaml

+51
Original file line numberDiff line numberDiff line change
@@ -0,0 +1,51 @@
1+
#DATA
2+
data_root: /mnt/workspace/dongbo/ais/models/models/NEW/hhe/data/as/
3+
classes: 2
4+
modality: mri
5+
6+
7+
#MODEL
8+
arch: attention_unet
9+
in_chan: 1
10+
base_chan: 32
11+
down_scale: [[1,2,2], [1,2,2], [2,2,2], [2,2,2]]
12+
kernel_size: [[1,3,3], [1,3,3], [3,3,3], [3,3,3], [3,3,3]]
13+
block: BasicBlock
14+
norm: in
15+
16+
#TRAIN
17+
epochs: 150
18+
training_size: [32, 32, 80] # training crop size
19+
start_epoch: 0
20+
21+
22+
seed: 0
23+
k_fold: 5
24+
25+
optimizer: adamw
26+
base_lr: 0.001
27+
betas: [0.9, 0.999]
28+
weight_decay: 0.05 # weight decay of SGD optimizer
29+
weight: [0.5, 1, 1, 1] # weitght of each class in the loss function
30+
rlt: 1 # relation between CE and Dice loss
31+
32+
scale: [0.1, 0.3, 0.3] # scale for data augmentation 0.1 0.3 0.3
33+
rotate: [30, 0, 0] # rotation angle for data augmentation
34+
translate: [0, 0, 0]
35+
gaussian_noise_std: 0.02
36+
additive_brightness_std: 0.7
37+
gamma_range: [0.5, 1.6]
38+
39+
40+
iter_per_epoch: 200
41+
42+
43+
#VALIDATION
44+
ema: True
45+
ema_alpha: 0.99
46+
val_frequency: 1
47+
48+
49+
#INFERENCE
50+
sliding_window: False
51+
window_size: [32, 32, 80]

Diff for: ‎H3/AmygdalaGo-BOLT/config/acdc/daunet_2d.yaml

+43
Original file line numberDiff line numberDiff line change
@@ -0,0 +1,43 @@
1+
#DATA
2+
data_root: dataset/acdc_2d
3+
classes: 4
4+
modality: mri
5+
6+
7+
#MODEL
8+
arch: daunet
9+
in_chan: 1
10+
base_chan: 32
11+
block: BasicBlock
12+
13+
14+
#TRAIN
15+
epochs: 150
16+
training_size: [256, 256] # training crop size
17+
start_epoch: 0
18+
19+
seed: 0
20+
k_fold: 5
21+
22+
optimizer: adamw
23+
base_lr: 0.0005
24+
betas: [0.9, 0.999]
25+
weight_decay: 0.05 # weight decay of SGD optimizer
26+
weight: [0.5, 1, 1, 1] # weitght of each class in the loss function
27+
rlt: 1 # relation between CE and Dice loss
28+
29+
scale: 0.3 # scale for data augmentation
30+
rotate: 30 # rotation angle for data augmentation
31+
translate: 0
32+
gaussian_noise_std: 0.02
33+
additive_brightness_std: 0.1
34+
gamma_range: [0.5, 1.6]
35+
36+
37+
#VALIDATION
38+
ema: True
39+
ema_alpha: 0.99
40+
val_frequency: 10
41+
42+
#INFERENCE
43+
sliding_window: False

Diff for: ‎H3/AmygdalaGo-BOLT/config/acdc/readme.txt

+2
Original file line numberDiff line numberDiff line change
@@ -0,0 +1,2 @@
1+
1. 修改class和class weight.
2+
2. 修改预训练权重和数据集路径。

Diff for: ‎H3/AmygdalaGo-BOLT/config/acdc/resunet_2d.yaml

+44
Original file line numberDiff line numberDiff line change
@@ -0,0 +1,44 @@
1+
#DATA
2+
data_root: dataset/acdc_2d
3+
classes: 4
4+
modality: mri
5+
6+
7+
#MODEL
8+
arch: resunet
9+
in_chan: 1
10+
base_chan: 32
11+
block: BasicBlock
12+
13+
14+
#TRAIN
15+
epochs: 150
16+
training_size: [256, 256] # training crop size
17+
start_epoch: 0
18+
19+
seed: 0 # random seed for suffule before setting cross validation fold
20+
k_fold: 5 # number of folds in cross validation
21+
22+
optimizer: adamw
23+
base_lr: 0.001
24+
betas: [0.9, 0.999]
25+
#momentum: 0.9 # momentum of SGD optimizer
26+
weight_decay: 0.05 # weight decay of SGD optimizer
27+
weight: [0.5, 1, 1, 1] # weitght of each class in the loss function
28+
rlt: 1 # relation between CE and Dice loss
29+
30+
scale: 0.3 # scale for data augmentation
31+
rotate: 180 # rotation angle for data augmentation
32+
translate: 0
33+
gaussian_noise_std: 0.02
34+
additive_brightness_std: 0.7
35+
gamma_range: [0.5, 1.6]
36+
37+
38+
#VALIDATION
39+
ema: True
40+
ema_alpha: 0.99
41+
val_frequency: 10
42+
43+
#INFERENCE
44+
sliding_window: False

Diff for: ‎H3/AmygdalaGo-BOLT/config/acdc/resunet_3d.yaml

+52
Original file line numberDiff line numberDiff line change
@@ -0,0 +1,52 @@
1+
#DATA
2+
#DATA
3+
data_root: /mnt/workspace/dongbo/ais/models/models/NEW/hhe/data/as/
4+
classes: 2
5+
modality: mri
6+
7+
8+
#MODEL
9+
arch: resunet
10+
in_chan: 1
11+
base_chan: 32
12+
down_scale: [[1,2,2], [1,2,2], [2,2,2], [2,2,2]]
13+
kernel_size: [[1,3,3], [1,3,3], [3,3,3], [3,3,3], [3,3,3]]
14+
block: BasicBlock
15+
norm: in
16+
17+
#TRAIN
18+
epochs: 8
19+
training_size: [32, 32, 80] # training crop size
20+
start_epoch: 0
21+
22+
seed: 0
23+
k_fold: 1
24+
25+
optimizer: adamw
26+
base_lr: 0.001
27+
betas: [0.9, 0.999]
28+
weight_decay: 0.05
29+
weight: [0.5, 1] # weitght of each class in the loss function
30+
rlt: 1 # relation between CE and Dice loss
31+
32+
scale: [0.1, 0.3, 0.3] # scale for data augmentation 0.1 0.3 0.3
33+
rotate: [30, 0, 0] # rotation angle for data augmentation
34+
translate: [0, 0, 0]
35+
gaussian_noise_std: 0.02
36+
additive_brightness_std: 0.7
37+
gamma_range: [0.5, 1.6]
38+
39+
40+
iter_per_epoch: 200
41+
42+
43+
#VALIDATION
44+
ema: True
45+
ema_alpha: 0.99
46+
val_frequency: 1
47+
48+
49+
50+
#INFERENCE
51+
sliding_window: False
52+
window_size: [32, 32, 80]

Diff for: ‎H3/AmygdalaGo-BOLT/config/acdc/swinunet_2d.yaml

+41
Original file line numberDiff line numberDiff line change
@@ -0,0 +1,41 @@
1+
#DATA
2+
data_root: dataset/acdc_2d
3+
classes: 4
4+
modality: mri
5+
6+
7+
#MODEL
8+
arch: swinunet
9+
init_model: '/research/cbim/vast/yg397/github/UTNet/initmodel/swin_tiny_patch4_window7_224.pth'
10+
11+
#TRAIN
12+
epochs: 400
13+
training_size: [224, 224] # training crop size
14+
start_epoch: 0
15+
16+
seed: 0
17+
k_fold: 5
18+
19+
20+
optimizer: adamw
21+
base_lr: 0.0005
22+
betas: [0.9, 0.999]
23+
weight_decay: 0.05 # weight decay of SGD optimizer
24+
weight: [0.5, 1, 1, 1] # weitght of each class in the loss function
25+
rlt: 1 # relation between CE and Dice loss
26+
27+
scale: 0.3 # scale for data augmentation
28+
rotate: 180 # rotation angle for data augmentation
29+
translate: 0
30+
gaussian_noise_std: 0.02
31+
additive_brightness_std: 0.7
32+
gamma_range: [0.5, 1.6]
33+
34+
35+
#VALIDATION
36+
ema: True
37+
ema_alpha: 0.99
38+
val_frequency: 10
39+
40+
#INFERENCE
41+
sliding_window: False

Diff for: ‎H3/AmygdalaGo-BOLT/config/acdc/transunet_2d.yaml

+40
Original file line numberDiff line numberDiff line change
@@ -0,0 +1,40 @@
1+
#DATA
2+
data_root: dataset/acdc_2d
3+
classes: 4
4+
modality: mri
5+
6+
7+
#MODEL
8+
arch: transunet
9+
init_model: '/research/cbim/vast/yg397/github/UTNet/initmodel/R50+ViT-B_16.npz'
10+
11+
#TRAIN
12+
epochs: 150
13+
training_size: [256, 256] # training crop size
14+
start_epoch: 0
15+
16+
seed: 0
17+
k_fold: 5
18+
19+
optimizer: adamw
20+
base_lr: 0.0005
21+
betas: [0.9, 0.999]
22+
weight_decay: 0.05
23+
weight: [0.5, 1, 1, 1] # weitght of each class in the loss function
24+
rlt: 1 # relation between CE and Dice loss
25+
26+
scale: 0.3 # scale for data augmentation
27+
rotate: 180 # rotation angle for data augmentation
28+
translate: 0
29+
gaussian_noise_std: 0.02
30+
additive_brightness_std: 0.7
31+
gamma_range: [0.5, 1.6]
32+
33+
34+
#VALIDATION
35+
ema: True
36+
ema_alpha: 0.99
37+
val_frequency: 10
38+
39+
#INFERENCE
40+
sliding_window: False

Diff for: ‎H3/AmygdalaGo-BOLT/config/acdc/unet++_2d.yaml

+44
Original file line numberDiff line numberDiff line change
@@ -0,0 +1,44 @@
1+
#DATA
2+
data_root: dataset/acdc_2d
3+
classes: 4
4+
modality: mri
5+
6+
7+
#MODEL
8+
arch: unet++
9+
in_chan: 1
10+
base_chan: 32
11+
12+
13+
#TRAIN
14+
epochs: 150
15+
training_size: [256, 256] # training crop size
16+
start_epoch: 0
17+
18+
seed: 0
19+
k_fold: 5
20+
21+
optimizer: adamw
22+
base_lr: 0.0005
23+
betas: [0.9, 0.999]
24+
weight_decay: 0.05 # weight decay of SGD optimizer
25+
weight: [0.5, 1, 1, 1] # weitght of each class in the loss function
26+
rlt: 1 # relation between CE and Dice loss
27+
28+
scale: 0.3 # scale for data augmentation
29+
rotate: 180 # rotation angle for data augmentation
30+
translate: 0
31+
gaussian_noise_std: 0.02
32+
additive_brightness_std: 0.7
33+
gamma_range: [0.5, 1.6]
34+
35+
36+
#VALIDATION
37+
ema: True
38+
ema_alpha: 0.99
39+
val_frequency: 10
40+
41+
42+
43+
#INFERENCE
44+
sliding_window: False

0 commit comments

Comments
 (0)