-
Notifications
You must be signed in to change notification settings - Fork 2
/
Copy pathconfig.yml
116 lines (95 loc) · 3.85 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
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
# Configuration
# -------- Directories --------
# Enumerates the directories to the files needed for the large scale detection
# Directories :
# - path/to/source/images : the path to the source images
# - path/to/topo : the path to the BDTOPO
# - path/to/iris : the path to the IRIS dataset, used in the postprocessing step.
# - path/to/communes : the path to the geographic information of the communes, used in the
# postprocessing step.
# - path/to/model : the path to the model that will be used for detection.
#
# Custom directories
# - temp : the name of the directory that stores the intermediary outputs
# - aux : the directory that contains auxiliary files (e.g. buildings)
#
# - data : the name of the directory in which the final output (.geojson file)
# are stored
source_images_dir : ""
model_dir : ''
source_commune_dir : ''
source_topo_dir : ""
temp_dir : 'temp'
aux_dir: 'aux'
outputs_dir : 'data'
# -------- Parameters --------
# Parameters for the main pipeline script
# - run_xx.py : True if one want to run specific parts of the pipeline, False otherwise.
run_classification : True
run_segmentation : True
run_aggregation : True
# Parameters that are common to certain pipelines:
# Patch_size : used in the pre processing and the detection steps.
# - departement_number : the departement number, used in the postprocessing and main script.
patch_size : 224
step_size : 137 # the step size (in pixels) between two thumbnail centers
device : 'cuda'
# Parameters that are specific to certain pipelines
# Preprocessing
# tiles_list : the list of tiles to proceed. By default it is set to None, but it is edited in the
# tutorial to speed-up the execution of the tutorial
tiles_list :
- 69-2020-0815-6525-LA93-0M20-E080
- 69-2020-0810-6535-LA93-0M20-E080
- 69-2020-0830-6535-LA93-0M20-E080
- 69-2020-0825-6525-LA93-0M20-E080
- 69-2020-0820-6535-LA93-0M20-E080
- 69-2020-0835-6525-LA93-0M20-E080
- 69-2020-0810-6530-LA93-0M20-E080
- 69-2020-0815-6520-LA93-0M20-E080
- 69-2020-0820-6530-LA93-0M20-E080
- 69-2020-0835-6520-LA93-0M20-E080
- 69-2020-0830-6530-LA93-0M20-E080
- 69-2020-0825-6520-LA93-0M20-E080
- 69-2020-0820-6525-LA93-0M20-E080
- 69-2020-0835-6535-LA93-0M20-E080
- 69-2020-0830-6525-LA93-0M20-E080
- 69-2020-0825-6535-LA93-0M20-E080
- 69-2020-0810-6525-LA93-0M20-E080
- 69-2020-0815-6535-LA93-0M20-E080
- 69-2020-0830-6520-LA93-0M20-E080
- 69-2020-0825-6530-LA93-0M20-E080
- 69-2020-0820-6520-LA93-0M20-E080
- 69-2020-0835-6530-LA93-0M20-E080
- 69-2020-0815-6530-LA93-0M20-E080
- 69-2020-0810-6520-LA93-0M20-E080
# Classification
# - device : the device to use. If None, the device is automatically set up
# - cls_batch_size : the batch_size
# - cls_threshold : the value of the thresold. Should be deliberatily low because false
# positives (that are not located on buildings) are filtered out
# - cls_model : the name of the model to look for in the folder.
cls_batch_size : 256
cls_threshold : 0.51
cls_model : "model_oracle"
# Segmentation
# - seg_batch_size : the size of the segmentation batch size
# - seg_threshold : the threshold chosen for segmentation. Should be
# the value that maximizes performance in order to get
# masks that are as accurate as possible
# - num_gpu : the number of GPU devices to use for inference
# - seg_model : the name of the model to look for in the folder.
#
seg_threshold : 0.46
num_gpu : 1
seg_batch_size : 64
seg_model : 'model_bdappv_seg'
# Aggregation :
# - filter building : keep only the installations that are on a building
# - filter LUT : compute the tilt with the extrernal look up table
# - constant_kWp : constant linear fit between the surface and installed capacity
# - look_up_table_dir : the diectory where the look up table is located.
filter_building : False
filter_LUT : True
constant_kWp : False
look_up_table_dir : ""