File size: 9,732 Bytes
e2cba1f | 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 | # Copyright The PyTorch Lightning team.
#
# Licensed under the Apache License, Version 2.0 (the "License");
# you may not use this file except in compliance with the License.
# You may obtain a copy of the License at
#
# http://www.apache.org/licenses/LICENSE-2.0
#
# Unless required by applicable law or agreed to in writing, software
# distributed under the License is distributed on an "AS IS" BASIS,
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
# See the License for the specific language governing permissions and
# limitations under the License.
from typing import Any, Dict, List, Literal, Optional, Tuple, Union
import torch
from torch import Tensor
from torchmetrics.detection.helpers import (
CocoBackend,
_calculate_map_with_coco,
_get_safe_item_values,
_input_validator,
_validate_iou_type_arg,
)
from torchmetrics.utilities.imports import (
_FASTER_COCO_EVAL_AVAILABLE,
_PYCOCOTOOLS_AVAILABLE,
_TORCHVISION_AVAILABLE,
)
if not (_PYCOCOTOOLS_AVAILABLE or _FASTER_COCO_EVAL_AVAILABLE) or not _TORCHVISION_AVAILABLE:
__doctest_skip__ = [
"mean_average_precision",
]
def mean_average_precision(
preds: List[Dict[str, Any]],
target: List[Dict[str, Any]],
box_format: Literal["xyxy", "xywh", "cxcywh"] = "xyxy",
iou_type: Union[Literal["bbox", "segm"], Tuple[Literal["bbox", "segm"], ...]] = "bbox",
iou_thresholds: Optional[list[float]] = None,
rec_thresholds: Optional[list[float]] = None,
max_detection_thresholds: Optional[list[int]] = None,
class_metrics: bool = False,
extended_summary: bool = False,
average: Literal["macro", "micro"] = "macro",
backend: Literal["pycocotools", "faster_coco_eval"] = "pycocotools",
warn_on_many_detections: bool = True,
) -> Union[Tensor, Dict[str, Tensor]]:
r"""Compute the mean average precision (mAP) and mean average recall (mAR) for object detection predictions.
This function evaluates detection predictions for either bounding boxes or segmentation masks based
on the provided ``iou_type``, comparing predictions (``preds``) and ground truth annotations (``target``)
using a COCO-style evaluation. The expected input for each image is a dictionary with keys:
- For bounding boxes (``iou_type="bbox"``): ``boxes``, ``scores``, and ``labels``.
- For segmentation (``iou_type="segm"``): ``masks``, ``scores``, and ``labels``.
In addition, ground truth dictionaries may include the optional keys ``iscrowd`` and ``area``.
Boxes are expected in the coordinate format provided via ``box_format``, which supports:
- ``"xyxy"``: [xmin, ymin, xmax, ymax]
- ``"xywh"``: [xmin, ymin, width, height]
- ``"cxcywh"``: [center_x, center_y, width, height]
The evaluation defaults to IoU thresholds from 0.50 to 0.95 (step 0.05), recall thresholds
from 0.00 to 1.00 (step 0.01), and maximum detection thresholds of [1, 10, 100]. These can be overridden
by specifying ``iou_thresholds``, ``rec_thresholds``, and ``max_detection_thresholds``, respectively.
Optionally, per-class metrics may be computed by enabling ``class_metrics``, and an extended summary
(including IoU, precision, recall, and scores) is available via ``extended_summary``.
The averaging method over labels can be set with ``average`` ("macro" or "micro") and the evaluation
is performed using either the ``pycocotools`` or ``faster_coco_eval`` backend.
Args:
preds: List of dictionaries, each representing detection predictions for a single image.
target: List of dictionaries, each representing ground truth annotations for a single image.
box_format: Format of the input bounding boxes. Supported values are "xyxy", "xywh", and "cxcywh".
iou_type: Type of IoU to compute. Can be "bbox", "segm", or a tuple containing both.
iou_thresholds: List of IoU thresholds (default is [0.5, 0.55, ..., 0.95]).
rec_thresholds: List of recall thresholds (default is [0.0, 0.01, ..., 1.0]).
max_detection_thresholds: List of maximum detections per image (default is [1, 10, 100]).
class_metrics: Whether to compute per-class mAP and mAR metrics.
extended_summary: Whether to include additional outputs (IoU, precision, recall, scores) in the result.
average: Averaging method over labels, either "macro" or "micro".
backend: Backend to use for evaluation ("pycocotools" or "faster_coco_eval").
warn_on_many_detections: If True, warn when there are an unusually large number of detections.
Returns:
dict: A dictionary containing the evaluation metrics. The dictionary includes the following keys:
- ``map``: Global mean average precision over the defined IoU thresholds.
- ``mar_{max_det}``: Global mean average recall for each maximum detection threshold.
- ``map_per_class``: Mean average precision per observed class (or -1 if ``class_metrics`` is disabled).
- ``mar_{max_det}_per_class``: Mean average recall per observed class for the highest detection threshold.
- ``classes``: A tensor listing all observed classes.
Example::
# Example with bounding boxes
>>> from torch import tensor
>>> from torchmetrics.functional.detection.map import mean_average_precision
>>> preds = [
... {
... "boxes": tensor([[258.0, 41.0, 606.0, 285.0]]),
... "scores": tensor([0.536]),
... "labels": tensor([0]),
... }
... ]
>>> target = [
... {
... "boxes": tensor([[214.0, 41.0, 562.0, 285.0]]),
... "labels": tensor([0]),
... }
... ]
>>> result = mean_average_precision(preds, target, iou_type="bbox")
>>> print(f"mAP: {result['map']:.4f}, mAP@0.5: {result['map_50']:.4f}")
mAP: 0.6000, mAP@0.5: 1.0000
Example::
# Example with segmentation masks
>>> import torch
>>> from torch import tensor
>>> from torchmetrics.functional.detection.map import mean_average_precision
>>> mask_pred = tensor([
... [0, 0, 0, 0, 0],
... [0, 0, 1, 1, 0],
... [0, 0, 1, 1, 0],
... [0, 0, 0, 0, 0],
... [0, 0, 0, 0, 0],
... ], dtype=torch.bool)
>>> mask_tgt = tensor([
... [0, 0, 0, 0, 0],
... [0, 0, 1, 0, 0],
... [0, 0, 1, 1, 0],
... [0, 0, 1, 0, 0],
... [0, 0, 0, 0, 0],
... ], dtype=torch.bool)
>>> preds = [
... {
... "masks": mask_pred.unsqueeze(0),
... "scores": tensor([0.536]),
... "labels": tensor([0]),
... }
... ]
>>> target = [
... {
... "masks": mask_tgt.unsqueeze(0),
... "labels": tensor([0]),
... }
... ]
>>> result = mean_average_precision(preds, target, iou_type="segm")
>>> print(f"mAP: {result['map']:.4f}, mAP@0.5: {result['map_50']:.4f}")
mAP: 0.2000, mAP@0.5: 1.0000
"""
iou_thresholds = iou_thresholds or torch.linspace(0.5, 0.95, round((0.95 - 0.5) / 0.05) + 1).tolist()
max_detection_thresholds = torch.sort(torch.tensor(max_detection_thresholds or [1, 10, 100], dtype=torch.int))[
0
].tolist()
rec_thresholds = rec_thresholds or torch.linspace(0.0, 1.00, round(1.00 / 0.01) + 1).tolist()
iou_type = _validate_iou_type_arg(iou_type)
_input_validator(preds, target, iou_type=iou_type)
coco_backend = CocoBackend(backend=backend)
detection_box: List[Tensor] = []
detection_labels: List[Tensor] = []
detection_scores: List[Tensor] = []
detection_mask: List[Tensor] = []
for item in preds:
bbox_detection, mask_detection = _get_safe_item_values(
iou_type, box_format, max_detection_thresholds, coco_backend, item, warn=warn_on_many_detections
)
if bbox_detection is not None:
detection_box.append(bbox_detection)
if mask_detection is not None:
detection_mask.append(mask_detection) # type: ignore[arg-type]
detection_labels.append(item["labels"])
detection_scores.append(item["scores"])
groundtruth_box: List[Tensor] = []
groundtruth_mask: List[Tensor] = []
groundtruth_labels: List[Tensor] = []
groundtruth_crowds: List[Tensor] = []
groundtruth_area: List[Tensor] = []
for item in target:
bbox_groundtruth, mask_groundtruth = _get_safe_item_values(
iou_type, box_format, max_detection_thresholds, coco_backend, item
)
if bbox_groundtruth is not None:
groundtruth_box.append(bbox_groundtruth)
if mask_groundtruth is not None:
groundtruth_mask.append(mask_groundtruth) # type: ignore[arg-type]
groundtruth_labels.append(item["labels"])
groundtruth_crowds.append(item.get("iscrowd", torch.zeros_like(item["labels"])))
groundtruth_area.append(item.get("area", torch.zeros_like(item["labels"])))
result_dict = _calculate_map_with_coco(
coco_backend,
groundtruth_labels,
groundtruth_box,
groundtruth_mask,
groundtruth_crowds,
groundtruth_area,
detection_labels,
detection_box,
detection_mask,
detection_scores,
iou_type,
average,
iou_thresholds,
rec_thresholds,
max_detection_thresholds,
class_metrics,
extended_summary,
)
return {k: (v.squeeze() if isinstance(v, torch.Tensor) and v.numel() == 1 else v) for k, v in result_dict.items()}
|