Spaces:
Running
on
Zero
Running
on
Zero
File size: 7,302 Bytes
2eafbc4 |
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 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 |
import random
from functools import partial
from typing import Any, Dict, Optional, Set
import numpy as np
from inference.core.active_learning.entities import (
Prediction,
PredictionType,
SamplingMethod,
)
from inference.core.constants import (
CLASSIFICATION_TASK,
INSTANCE_SEGMENTATION_TASK,
KEYPOINTS_DETECTION_TASK,
OBJECT_DETECTION_TASK,
)
from inference.core.exceptions import ActiveLearningConfigurationError
ELIGIBLE_PREDICTION_TYPES = {
CLASSIFICATION_TASK,
INSTANCE_SEGMENTATION_TASK,
KEYPOINTS_DETECTION_TASK,
OBJECT_DETECTION_TASK,
}
def initialize_close_to_threshold_sampling(
strategy_config: Dict[str, Any]
) -> SamplingMethod:
try:
selected_class_names = strategy_config.get("selected_class_names")
if selected_class_names is not None:
selected_class_names = set(selected_class_names)
sample_function = partial(
sample_close_to_threshold,
selected_class_names=selected_class_names,
threshold=strategy_config["threshold"],
epsilon=strategy_config["epsilon"],
only_top_classes=strategy_config.get("only_top_classes", True),
minimum_objects_close_to_threshold=strategy_config.get(
"minimum_objects_close_to_threshold",
1,
),
probability=strategy_config["probability"],
)
return SamplingMethod(
name=strategy_config["name"],
sample=sample_function,
)
except KeyError as error:
raise ActiveLearningConfigurationError(
f"In configuration of `close_to_threshold_sampling` missing key detected: {error}."
) from error
def sample_close_to_threshold(
image: np.ndarray,
prediction: Prediction,
prediction_type: PredictionType,
selected_class_names: Optional[Set[str]],
threshold: float,
epsilon: float,
only_top_classes: bool,
minimum_objects_close_to_threshold: int,
probability: float,
) -> bool:
if is_prediction_a_stub(prediction=prediction):
return False
if prediction_type not in ELIGIBLE_PREDICTION_TYPES:
return False
close_to_threshold = prediction_is_close_to_threshold(
prediction=prediction,
prediction_type=prediction_type,
selected_class_names=selected_class_names,
threshold=threshold,
epsilon=epsilon,
only_top_classes=only_top_classes,
minimum_objects_close_to_threshold=minimum_objects_close_to_threshold,
)
if not close_to_threshold:
return False
return random.random() < probability
def is_prediction_a_stub(prediction: Prediction) -> bool:
return prediction.get("is_stub", False)
def prediction_is_close_to_threshold(
prediction: Prediction,
prediction_type: PredictionType,
selected_class_names: Optional[Set[str]],
threshold: float,
epsilon: float,
only_top_classes: bool,
minimum_objects_close_to_threshold: int,
) -> bool:
if CLASSIFICATION_TASK not in prediction_type:
return detections_are_close_to_threshold(
prediction=prediction,
selected_class_names=selected_class_names,
threshold=threshold,
epsilon=epsilon,
minimum_objects_close_to_threshold=minimum_objects_close_to_threshold,
)
checker = multi_label_classification_prediction_is_close_to_threshold
if "top" in prediction:
checker = multi_class_classification_prediction_is_close_to_threshold
return checker(
prediction=prediction,
selected_class_names=selected_class_names,
threshold=threshold,
epsilon=epsilon,
only_top_classes=only_top_classes,
)
def multi_class_classification_prediction_is_close_to_threshold(
prediction: Prediction,
selected_class_names: Optional[Set[str]],
threshold: float,
epsilon: float,
only_top_classes: bool,
) -> bool:
if only_top_classes:
return (
multi_class_classification_prediction_is_close_to_threshold_for_top_class(
prediction=prediction,
selected_class_names=selected_class_names,
threshold=threshold,
epsilon=epsilon,
)
)
for prediction_details in prediction["predictions"]:
if class_to_be_excluded(
class_name=prediction_details["class"],
selected_class_names=selected_class_names,
):
continue
if is_close_to_threshold(
value=prediction_details["confidence"], threshold=threshold, epsilon=epsilon
):
return True
return False
def multi_class_classification_prediction_is_close_to_threshold_for_top_class(
prediction: Prediction,
selected_class_names: Optional[Set[str]],
threshold: float,
epsilon: float,
) -> bool:
if (
selected_class_names is not None
and prediction["top"] not in selected_class_names
):
return False
return abs(prediction["confidence"] - threshold) < epsilon
def multi_label_classification_prediction_is_close_to_threshold(
prediction: Prediction,
selected_class_names: Optional[Set[str]],
threshold: float,
epsilon: float,
only_top_classes: bool,
) -> bool:
predicted_classes = set(prediction["predicted_classes"])
for class_name, prediction_details in prediction["predictions"].items():
if only_top_classes and class_name not in predicted_classes:
continue
if class_to_be_excluded(
class_name=class_name, selected_class_names=selected_class_names
):
continue
if is_close_to_threshold(
value=prediction_details["confidence"], threshold=threshold, epsilon=epsilon
):
return True
return False
def detections_are_close_to_threshold(
prediction: Prediction,
selected_class_names: Optional[Set[str]],
threshold: float,
epsilon: float,
minimum_objects_close_to_threshold: int,
) -> bool:
detections_close_to_threshold = count_detections_close_to_threshold(
prediction=prediction,
selected_class_names=selected_class_names,
threshold=threshold,
epsilon=epsilon,
)
return detections_close_to_threshold >= minimum_objects_close_to_threshold
def count_detections_close_to_threshold(
prediction: Prediction,
selected_class_names: Optional[Set[str]],
threshold: float,
epsilon: float,
) -> int:
counter = 0
for prediction_details in prediction["predictions"]:
if class_to_be_excluded(
class_name=prediction_details["class"],
selected_class_names=selected_class_names,
):
continue
if is_close_to_threshold(
value=prediction_details["confidence"], threshold=threshold, epsilon=epsilon
):
counter += 1
return counter
def class_to_be_excluded(
class_name: str, selected_class_names: Optional[Set[str]]
) -> bool:
return selected_class_names is not None and class_name not in selected_class_names
def is_close_to_threshold(value: float, threshold: float, epsilon: float) -> bool:
return abs(value - threshold) < epsilon
|