"""Image feaures."""
from typing import Any
import torch
from ...dataset.base import LAMLDataset
from ...image.utils import pil_loader
from ...transformers.base import ColumnsSelector
from ...transformers.base import LAMLTransformer
from ...transformers.base import SequentialTransformer
from ...transformers.base import UnionTransformer
from ...transformers.image import AutoCVWrap
from ...transformers.image import ImageFeaturesTransformer
from ...transformers.numeric import FillInf
from ...transformers.numeric import FillnaMedian
from ...transformers.numeric import StandardScaler
from ..utils import get_columns_by_role
from .base import FeaturesPipeline
[docs]class ImageDataFeatures:
"""Class contains basic features transformations for image data.
Args:
**kwargs: Default parameters.
"""
def __init__(self, **kwargs: Any):
self.hist_size = 30
self.is_hsv = True
self.n_jobs = 4
self.loader = pil_loader
self.embed_model = "efficientnet_b0.ra_in1k"
self.weights_path = None
self.subs = 10000
self.cache_dir = "../cache_CV"
self.device = torch.device("cuda:0") if torch.cuda.is_available() else torch.device("cpu")
self.batch_size = 128
self.verbose = True
self.random_state = 42
for k in kwargs:
self.__dict__[k] = kwargs[k]
[docs]class ImageSimpleFeatures(FeaturesPipeline, ImageDataFeatures):
"""Class contains simple color histogram features for image data."""
[docs] def create_pipeline(self, train: LAMLDataset) -> LAMLTransformer:
"""Create pipeline for images data.
Args:
train: Dataset with train features.
Returns:
Transformer.
"""
transformers_list = []
# process texts
imgs = get_columns_by_role(train, "Path")
if len(imgs) > 0:
imgs_processing = SequentialTransformer(
[
ColumnsSelector(keys=imgs),
ImageFeaturesTransformer(self.hist_size, self.is_hsv, self.n_jobs, self.loader),
SequentialTransformer([FillInf(), FillnaMedian(), StandardScaler()]),
]
)
transformers_list.append(imgs_processing)
union_all = UnionTransformer(transformers_list)
return union_all
[docs]class ImageAutoFeatures(FeaturesPipeline, ImageDataFeatures):
"""Class contains efficient-net embeddings features for image data."""
[docs] def create_pipeline(self, train: LAMLDataset) -> LAMLTransformer:
"""Create pipeline for images data.
Args:
train: Dataset with train features.
Returns:
Transformer.
"""
transformers_list = []
# process texts
imgs = get_columns_by_role(train, "Path")
if len(imgs) > 0:
imgs_processing = SequentialTransformer(
[
ColumnsSelector(keys=imgs),
AutoCVWrap(
self.embed_model,
self.weights_path,
self.cache_dir,
self.subs,
self.device,
self.n_jobs,
self.random_state,
self.batch_size,
self.verbose,
),
SequentialTransformer([FillInf(), FillnaMedian(), StandardScaler()]),
]
)
transformers_list.append(imgs_processing)
union_all = UnionTransformer(transformers_list)
return union_all