-
Notifications
You must be signed in to change notification settings - Fork 1
/
Copy pathtrain_script.sh
executable file
·94 lines (92 loc) · 2.85 KB
/
train_script.sh
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
#!/bin/bash
MODEL_NAME='din50'
TARGET_SET='webpage'
TARGET_IMAGE='/path/to/target/dataset/image'
TARGET_FIXATION='/path/to/target/dataset/fixationmap'
SOURCE_SET='salicon'
SOURCE_IMAGE='/path/to/source/dataset/image'
SOURCE_FIXATION='/path/to/source/dataset/fixationmap'
OUTPUT_DIR='logs/'${SOURCE_SET}'_'${MODEL_NAME}'_'${TARGET_SET}
PRETRAINED_EPOCH=3
PRETRAINED="./model/pretrained_DINet.pth"
LR=0.00005
BATCH_SIZE=5 # 1 3 5 for 1-, 5-, and 10-shot
NUM_SHOTS=10
TRIALS=1
export CUDA_VISIBLE_DEVICES=0
# conventional training
python train.py \
--batch_size 10 \
--epochs 10 \
--lr ${LR} \
--lr_decay_epoch 3 \
--lr_coef 0.2 \
--weight_decay 1e-4 \
--model ${MODEL_NAME} \
--out_dir "${OUTPUT_DIR}/baseline/" \
--train_img_dir "${SOURCE_IMAGE}/" \
--train_gt_dir "${SOURCE_FIXATION}/" \
--val_img_dir "${TARGET_IMAGE}/" \
--val_gt_dir "${TARGET_FIXATION}/" \
--image_size 320 480 \
--tr_fxt_size 480 640 \
--val_fxt_size 1360 768
# n-reference learning
python train_nref.py \
--batch_size 10 \
--epochs 10 \
--lr ${LR} \
--lr_decay_epoch 3 \
--lr_coef 0.2 \
--weight_decay 1e-4 \
--model ${MODEL_NAME} \
--out_dir "${OUTPUT_DIR}/${NUM_SHOTS}shot_${TRIALS}/tr_ref_val" \
--train_img_dir "${SOURCE_IMAGE}/" \
--train_gt_dir "${SOURCE_FIXATION}/" \
--val_img_dir "${TARGET_IMAGE}/" \
--val_gt_dir "${TARGET_FIXATION}/" \
--num_shots ${NUM_SHOTS} \
--ref_batch_size ${BATCH_SIZE} \
--image_size 320 480 \
--tr_fxt_size 480 640 \
--val_fxt_size 1360 768 \
--ref_fxt_size 1360 768
# n-shot learning
python train_nshot.py \
--batch_size 10 \
--epochs 10 \
--lr ${LR} \
--lr_decay_epoch 3 \
--lr_coef 0.2 \
--weight_decay 1e-4 \
--model ${MODEL_NAME} \
--out_dir "${OUTPUT_DIR}/${NUM_SHOTS}shot_${TRIALS}/tr_ft_val" \
--train_img_dir "${SOURCE_IMAGE}/" \
--train_gt_dir "${SOURCE_FIXATION}/" \
--val_img_dir "${TARGET_IMAGE}/" \
--val_gt_dir "${TARGET_FIXATION}/" \
--pretrainedModel ${PRETRAINED} \
--split_file "${OUTPUT_DIR}/${NUM_SHOTS}shot_${TRIALS}/tr_ref_val/split_data.npz" \
--num_shots ${NUM_SHOTS} \
--image_size 320 480 \
--val_fxt_size 1360 768
# n-reference fine-tuning
python train_nshot.py \
--batch_size 10 \
--epochs 10 \
--lr ${LR} \
--lr_decay_epoch 3 \
--lr_coef 0.2 \
--weight_decay 1e-4 \
--model ${MODEL_NAME} \
--out_dir "${OUTPUT_DIR}/${NUM_SHOTS}shot_${TRIALS}/tr_ref_ft_val" \
--train_img_dir "${SOURCE_IMAGE}/" \
--train_gt_dir "${SOURCE_FIXATION}/" \
--val_img_dir "${TARGET_IMAGE}/" \
--val_gt_dir "${TARGET_FIXATION}/" \
--pretrainedModel "${OUTPUT_DIR}/${NUM_SHOTS}shot_${TRIALS}/tr_ref_val/snapshots/model_ep${PRETRAINED_EPOCH}.pth" \
--pretrainedModel_head "${OUTPUT_DIR}/${NUM_SHOTS}shot_${TRIALS}/tr_ref_val/snapshots/model_head_ep${PRETRAINED_EPOCH}.pth" \
--split_file "${OUTPUT_DIR}/${NUM_SHOTS}shot_${TRIALS}/tr_ref_val/split_data.npz" \
--num_shots ${NUM_SHOTS} \
--image_size 320 480 \
--val_fxt_size 1360 768