jonathan-roberts1
commited on
Commit
•
ba0402e
1
Parent(s):
04cbbba
Ensure image is first feature
Browse files
SATIN.py
CHANGED
@@ -7,11 +7,9 @@ _CONSTITUENT_DATASETS = ['SAT-4', 'SAT-6', 'NASC-TG2', 'WHU-RS19', 'RSSCN7', 'RS
|
|
7 |
'Airbus-Wind-Turbines-Patches', 'USTC_SmokeRS', 'Canadian_Cropland',
|
8 |
'Ships-In-Satellite-Imagery', 'Satellite-Images-of-Hurricane-Damage',
|
9 |
'Brazilian_Coffee_Scenes', 'Brazilian_Cerrado-Savanna_Scenes', 'Million-AID',
|
10 |
-
'UC_Merced_LandUse_MultiLabel', 'MLRSNet',
|
11 |
'MultiScene', 'RSI-CB256']
|
12 |
-
|
13 |
-
_CONSTITUENT_DATASETS = ['EuroSAT']
|
14 |
-
"""
|
15 |
|
16 |
class SATINConfig(datasets.BuilderConfig):
|
17 |
"""BuilderConfig for SATIN"""
|
@@ -24,38 +22,22 @@ class SATINConfig(datasets.BuilderConfig):
|
|
24 |
self.description = None
|
25 |
self.features = None
|
26 |
|
27 |
-
#stream_dataset_info = load_dataset(self.hf_dataset_name, streaming=True, split='train').info
|
28 |
-
#self.description = stream_dataset_info.description
|
29 |
-
#self.features = stream_dataset_info.features
|
30 |
-
|
31 |
|
32 |
class SATIN(datasets.GeneratorBasedBuilder):
|
33 |
"""SATIN Images dataset"""
|
34 |
|
35 |
BUILDER_CONFIGS = [SATINConfig(name=dataset_name) for dataset_name in _CONSTITUENT_DATASETS]
|
36 |
|
37 |
-
"""
|
38 |
-
def __init__(self, *args, **kwargs):
|
39 |
-
super().__init__(*args, **kwargs)
|
40 |
-
self.config.hf_dataset = load_dataset(self.config.hf_dataset_name)
|
41 |
-
self.config.description = self.config.hf_dataset['train'].description
|
42 |
-
self.config.features = self.config.hf_dataset['train'].features
|
43 |
-
print(self.config.features)
|
44 |
-
"""
|
45 |
-
|
46 |
def _info(self):
|
47 |
if self.config.description is None or self.config.features is None:
|
48 |
stream_dataset_info = load_dataset(self.config.hf_dataset_name, streaming=True, split='train').info
|
49 |
self.config.description = stream_dataset_info.description
|
50 |
self.config.features = stream_dataset_info.features
|
51 |
-
#print(f'info {self.config.features}')
|
52 |
return datasets.DatasetInfo(
|
53 |
description=self.config.description,
|
54 |
features=self.config.features,
|
55 |
-
#supervised_keys=("image", "label"),
|
56 |
)
|
57 |
|
58 |
-
|
59 |
def _split_generators(self, dl_manager):
|
60 |
dataset = load_dataset(self.config.hf_dataset_name)
|
61 |
return [
|
@@ -71,7 +53,11 @@ class SATIN(datasets.GeneratorBasedBuilder):
|
|
71 |
huggingface_dataset = data_path['train']
|
72 |
features = huggingface_dataset.features
|
73 |
for idx, row in enumerate(huggingface_dataset):
|
74 |
-
|
|
|
|
|
|
|
|
|
75 |
|
76 |
|
77 |
|
|
|
7 |
'Airbus-Wind-Turbines-Patches', 'USTC_SmokeRS', 'Canadian_Cropland',
|
8 |
'Ships-In-Satellite-Imagery', 'Satellite-Images-of-Hurricane-Damage',
|
9 |
'Brazilian_Coffee_Scenes', 'Brazilian_Cerrado-Savanna_Scenes', 'Million-AID',
|
10 |
+
'UC_Merced_LandUse_MultiLabel', 'MLRSNet',
|
11 |
'MultiScene', 'RSI-CB256']
|
12 |
+
|
|
|
|
|
13 |
|
14 |
class SATINConfig(datasets.BuilderConfig):
|
15 |
"""BuilderConfig for SATIN"""
|
|
|
22 |
self.description = None
|
23 |
self.features = None
|
24 |
|
|
|
|
|
|
|
|
|
25 |
|
26 |
class SATIN(datasets.GeneratorBasedBuilder):
|
27 |
"""SATIN Images dataset"""
|
28 |
|
29 |
BUILDER_CONFIGS = [SATINConfig(name=dataset_name) for dataset_name in _CONSTITUENT_DATASETS]
|
30 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
31 |
def _info(self):
|
32 |
if self.config.description is None or self.config.features is None:
|
33 |
stream_dataset_info = load_dataset(self.config.hf_dataset_name, streaming=True, split='train').info
|
34 |
self.config.description = stream_dataset_info.description
|
35 |
self.config.features = stream_dataset_info.features
|
|
|
36 |
return datasets.DatasetInfo(
|
37 |
description=self.config.description,
|
38 |
features=self.config.features,
|
|
|
39 |
)
|
40 |
|
|
|
41 |
def _split_generators(self, dl_manager):
|
42 |
dataset = load_dataset(self.config.hf_dataset_name)
|
43 |
return [
|
|
|
53 |
huggingface_dataset = data_path['train']
|
54 |
features = huggingface_dataset.features
|
55 |
for idx, row in enumerate(huggingface_dataset):
|
56 |
+
features_dict = {feature: row[feature] for feature in features}
|
57 |
+
# Reorder features to make image the first feature
|
58 |
+
image = features_dict.pop('image')
|
59 |
+
features_dict = {'image': image, **features_dict}
|
60 |
+
yield idx, features_dict
|
61 |
|
62 |
|
63 |
|