-
Notifications
You must be signed in to change notification settings - Fork 39
/
config.yml
95 lines (78 loc) · 2.48 KB
/
config.yml
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
95
#Config
### Data Generation
#glob path to sensor data, recursive wildcards allowed
rgb_sensor_pool: /orange/ewhite/NeonData/OSBS/DP3.30010.001/**/Camera/**/*.tif
HSI_sensor_pool: /orange/ewhite/NeonData/OSBS/DP3.30006.001/**/Reflectance/*.h5
CHM_pool: /orange/ewhite/NeonData/**/CanopyHeightModelGtif/*.tif
HSI_tif_dir: /orange/idtrees-collab/Hyperspectral_tifs/year/
#NEON data filtering
min_stem_diameter: 10
#Minimum number of test samples per class to be included
min_train_samples: 10
min_test_samples: 10
#Iterations to search for the best train-test split that mantains the largest number of species. Set to 1 for debugging, no files will be written
iterations: 100
#minimum CHM height, leave blank to skip CHM filtering
min_CHM_height: 1
#Max difference between measured height and CHM height if CHM > height
max_CHM_diff: 4
#Max difference between measured height and CHM height if CHM < height
CHM_height_limit: 8
#Megaplot XY Data, these are .shp files saved as the siteID
megaplot_dir: /orange/idtrees-collab/megaplot/
#Crop generation, whether to make a new dataset and customize which parts to recreate
#Checkout data artifact from comet
use_data_commit: 67ec871c49cf472c8e1ae70b185addb1
#Make new dataset
data_dir: /blue/ewhite/b.weinstein/DeepTreeAttention/
convert_h5: True
#Overwrite existing crops
replace: True
#Use existing plotIDs for train/test split
new_train_test_split: True
#How many samples from non-osbs sites can we use per species
samples_from_other_sites: 0
oaks_sampling_ceiling: 200
other_sampling_ceiling: 500
evergreen_ceiling: 70
# Data loader
#resized Pixel size of the crowns. Square crops around each pixel of size x are used
image_size: 11
preload_images: True
workers: 20
sampling_ceiling: 200
#Network Parameters
pretrain_state_dict:
gpus: 1
batch_size: 128
bands: 349
#LR for each level x year
lr_0: 0.000001
lr_1: 0.000001
lr_2: 0.000005
lr_3: 0.0001
lr_4: 0.000005
fast_dev_run: False
accelerator: auto
epochs: 70
min_loss_weight: 10
#Evaluation config
#Top k class recall score
top_k: 4
#Number of individuals to plot in comet graphics tab
plot_n_individuals: 0
#Predict
predict_batch_size: 64
dead_threshold: 0.95
prediction_crop_dir: /blue/ewhite/b.weinstein/DeepTreeAttention/results/crops/
#Comet dashboard
comet_workspace: bw4sz
## Alive Dead Model Training
dead:
train_dir: data/raw/dead_train/
test_dir: data/raw/dead_test/
epochs: 40
lr: 0.0001
savedir: /orange/idtrees-collab/DeepTreeAttention/Dead/snapshots/
batch_size: 128
num_workers: 0