metrics.py 11.2 KB
Newer Older
W
weishengyu 已提交
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17
# copyright (c) 2021 PaddlePaddle Authors. All Rights Reserve.
#
# 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.

import numpy as np
import paddle
import paddle.nn as nn
C
cuicheng01 已提交
18 19 20 21 22 23
import paddle.nn.functional as F

from sklearn.metrics import hamming_loss
from sklearn.metrics import accuracy_score as accuracy_metric
from sklearn.metrics import multilabel_confusion_matrix
from sklearn.preprocessing import binarize
W
weishengyu 已提交
24

D
dongshuilong 已提交
25

W
weishengyu 已提交
26
class TopkAcc(nn.Layer):
W
weishengyu 已提交
27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43
    def __init__(self, topk=(1, 5)):
        super().__init__()
        assert isinstance(topk, (int, list, tuple))
        if isinstance(topk, int):
            topk = [topk]
        self.topk = topk

    def forward(self, x, label):
        if isinstance(x, dict):
            x = x["logits"]

        metric_dict = dict()
        for k in self.topk:
            metric_dict["top{}".format(k)] = paddle.metric.accuracy(
                x, label, k=k)
        return metric_dict

D
dongshuilong 已提交
44

W
weishengyu 已提交
45
class mAP(nn.Layer):
D
dongshuilong 已提交
46
    def __init__(self):
W
weishengyu 已提交
47 48
        super().__init__()

D
dongshuilong 已提交
49
    def forward(self, similarities_matrix, query_img_id, gallery_img_id,
D
dongshuilong 已提交
50
                keep_mask):
W
weishengyu 已提交
51
        metric_dict = dict()
D
dongshuilong 已提交
52 53 54 55 56 57 58 59 60 61 62

        choosen_indices = paddle.argsort(
            similarities_matrix, axis=1, descending=True)
        gallery_labels_transpose = paddle.transpose(gallery_img_id, [1, 0])
        gallery_labels_transpose = paddle.broadcast_to(
            gallery_labels_transpose,
            shape=[
                choosen_indices.shape[0], gallery_labels_transpose.shape[1]
            ])
        choosen_label = paddle.index_sample(gallery_labels_transpose,
                                            choosen_indices)
B
Bin Lu 已提交
63
        equal_flag = paddle.equal(choosen_label, query_img_id)
D
dongshuilong 已提交
64 65 66 67 68
        if keep_mask is not None:
            keep_mask = paddle.index_sample(
                keep_mask.astype('float32'), choosen_indices)
            equal_flag = paddle.logical_and(equal_flag,
                                            keep_mask.astype('bool'))
B
Bin Lu 已提交
69 70
        equal_flag = paddle.cast(equal_flag, 'float32')

D
dongshuilong 已提交
71 72 73 74 75 76
        num_rel = paddle.sum(equal_flag, axis=1)
        num_rel = paddle.greater_than(num_rel, paddle.to_tensor(0.))
        num_rel_index = paddle.nonzero(num_rel.astype("int"))
        num_rel_index = paddle.reshape(num_rel_index, [num_rel_index.shape[0]])
        equal_flag = paddle.index_select(equal_flag, num_rel_index, axis=0)

B
Bin Lu 已提交
77 78
        acc_sum = paddle.cumsum(equal_flag, axis=1)
        div = paddle.arange(acc_sum.shape[1]).astype("float32") + 1
D
dongshuilong 已提交
79
        precision = paddle.divide(acc_sum, div)
B
Bin Lu 已提交
80 81 82

        #calc map
        precision_mask = paddle.multiply(equal_flag, precision)
D
dongshuilong 已提交
83 84
        ap = paddle.sum(precision_mask, axis=1) / paddle.sum(equal_flag,
                                                             axis=1)
B
Bin Lu 已提交
85
        metric_dict["mAP"] = paddle.mean(ap).numpy()[0]
W
weishengyu 已提交
86 87
        return metric_dict

D
dongshuilong 已提交
88

W
weishengyu 已提交
89
class mINP(nn.Layer):
D
dongshuilong 已提交
90
    def __init__(self):
W
weishengyu 已提交
91 92
        super().__init__()

D
dongshuilong 已提交
93
    def forward(self, similarities_matrix, query_img_id, gallery_img_id,
D
dongshuilong 已提交
94
                keep_mask):
W
weishengyu 已提交
95
        metric_dict = dict()
D
dongshuilong 已提交
96 97 98 99 100 101 102 103 104 105 106

        choosen_indices = paddle.argsort(
            similarities_matrix, axis=1, descending=True)
        gallery_labels_transpose = paddle.transpose(gallery_img_id, [1, 0])
        gallery_labels_transpose = paddle.broadcast_to(
            gallery_labels_transpose,
            shape=[
                choosen_indices.shape[0], gallery_labels_transpose.shape[1]
            ])
        choosen_label = paddle.index_sample(gallery_labels_transpose,
                                            choosen_indices)
D
dongshuilong 已提交
107 108 109 110 111 112
        equal_flag = paddle.equal(choosen_label, query_img_id)
        if keep_mask is not None:
            keep_mask = paddle.index_sample(
                keep_mask.astype('float32'), choosen_indices)
            equal_flag = paddle.logical_and(equal_flag,
                                            keep_mask.astype('bool'))
D
dongshuilong 已提交
113
        equal_flag = paddle.cast(equal_flag, 'float32')
D
dongshuilong 已提交
114 115 116 117 118 119

        num_rel = paddle.sum(equal_flag, axis=1)
        num_rel = paddle.greater_than(num_rel, paddle.to_tensor(0.))
        num_rel_index = paddle.nonzero(num_rel.astype("int"))
        num_rel_index = paddle.reshape(num_rel_index, [num_rel_index.shape[0]])
        equal_flag = paddle.index_select(equal_flag, num_rel_index, axis=0)
B
Bin Lu 已提交
120 121

        #do accumulative sum
D
dongshuilong 已提交
122
        div = paddle.arange(equal_flag.shape[1]).astype("float32") + 2
D
dongshuilong 已提交
123 124
        minus = paddle.divide(equal_flag, div)
        auxilary = paddle.subtract(equal_flag, minus)
D
dongshuilong 已提交
125
        hard_index = paddle.argmax(auxilary, axis=1).astype("float32")
D
dongshuilong 已提交
126
        all_INP = paddle.divide(paddle.sum(equal_flag, axis=1), hard_index)
B
Bin Lu 已提交
127 128
        mINP = paddle.mean(all_INP)
        metric_dict["mINP"] = mINP.numpy()[0]
W
weishengyu 已提交
129 130
        return metric_dict

D
dongshuilong 已提交
131

W
weishengyu 已提交
132
class Recallk(nn.Layer):
D
dongshuilong 已提交
133
    def __init__(self, topk=(1, 5)):
W
weishengyu 已提交
134
        super().__init__()
B
Bin Lu 已提交
135
        assert isinstance(topk, (int, list, tuple))
W
weishengyu 已提交
136 137 138 139
        if isinstance(topk, int):
            topk = [topk]
        self.topk = topk

D
dongshuilong 已提交
140 141
    def forward(self, similarities_matrix, query_img_id, gallery_img_id,
                keep_mask):
W
weishengyu 已提交
142
        metric_dict = dict()
B
Bin Lu 已提交
143 144

        #get cmc
D
dongshuilong 已提交
145 146 147 148 149 150 151 152 153 154
        choosen_indices = paddle.argsort(
            similarities_matrix, axis=1, descending=True)
        gallery_labels_transpose = paddle.transpose(gallery_img_id, [1, 0])
        gallery_labels_transpose = paddle.broadcast_to(
            gallery_labels_transpose,
            shape=[
                choosen_indices.shape[0], gallery_labels_transpose.shape[1]
            ])
        choosen_label = paddle.index_sample(gallery_labels_transpose,
                                            choosen_indices)
B
Bin Lu 已提交
155
        equal_flag = paddle.equal(choosen_label, query_img_id)
D
dongshuilong 已提交
156 157 158 159 160
        if keep_mask is not None:
            keep_mask = paddle.index_sample(
                keep_mask.astype('float32'), choosen_indices)
            equal_flag = paddle.logical_and(equal_flag,
                                            keep_mask.astype('bool'))
B
Bin Lu 已提交
161
        equal_flag = paddle.cast(equal_flag, 'float32')
D
dongshuilong 已提交
162 163 164 165 166
        real_query_num = paddle.sum(equal_flag, axis=1)
        real_query_num = paddle.sum(
            paddle.greater_than(real_query_num, paddle.to_tensor(0.)).astype(
                "float32"))

B
Bin Lu 已提交
167
        acc_sum = paddle.cumsum(equal_flag, axis=1)
D
dongshuilong 已提交
168 169 170
        mask = paddle.greater_than(acc_sum,
                                   paddle.to_tensor(0.)).astype("float32")
        all_cmc = (paddle.sum(mask, axis=0) / real_query_num).numpy()
W
weishengyu 已提交
171 172 173 174 175

        for k in self.topk:
            metric_dict["recall{}".format(k)] = all_cmc[k - 1]
        return metric_dict

D
dongshuilong 已提交
176

B
Bin Lu 已提交
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
class Precisionk(nn.Layer):
    def __init__(self, topk=(1, 5)):
        super().__init__()
        assert isinstance(topk, (int, list, tuple))
        if isinstance(topk, int):
            topk = [topk]
        self.topk = topk

    def forward(self, similarities_matrix, query_img_id, gallery_img_id,
                keep_mask):
        metric_dict = dict()

        #get cmc
        choosen_indices = paddle.argsort(
            similarities_matrix, axis=1, descending=True)
        gallery_labels_transpose = paddle.transpose(gallery_img_id, [1, 0])
        gallery_labels_transpose = paddle.broadcast_to(
            gallery_labels_transpose,
            shape=[
                choosen_indices.shape[0], gallery_labels_transpose.shape[1]
            ])
        choosen_label = paddle.index_sample(gallery_labels_transpose,
                                            choosen_indices)
        equal_flag = paddle.equal(choosen_label, query_img_id)
        if keep_mask is not None:
            keep_mask = paddle.index_sample(
                keep_mask.astype('float32'), choosen_indices)
            equal_flag = paddle.logical_and(equal_flag,
                                            keep_mask.astype('bool'))
        equal_flag = paddle.cast(equal_flag, 'float32')
C
cuicheng01 已提交
207

B
Bin Lu 已提交
208 209 210 211 212 213 214 215 216 217
        Ns = paddle.arange(gallery_img_id.shape[0]) + 1
        equal_flag_cumsum = paddle.cumsum(equal_flag, axis=1)
        Precision_at_k = (paddle.mean(equal_flag_cumsum, axis=0) / Ns).numpy()

        for k in self.topk:
            metric_dict["precision@{}".format(k)] = Precision_at_k[k - 1]

        return metric_dict


218 219 220 221 222 223 224 225 226 227 228
class DistillationTopkAcc(TopkAcc):
    def __init__(self, model_key, feature_key=None, topk=(1, 5)):
        super().__init__(topk=topk)
        self.model_key = model_key
        self.feature_key = feature_key

    def forward(self, x, label):
        x = x[self.model_key]
        if self.feature_key is not None:
            x = x[self.feature_key]
        return super().forward(x, label)
C
cuicheng01 已提交
229 230 231 232 233 234 235 236 237 238 239 240


class GoogLeNetTopkAcc(TopkAcc):
    def __init__(self, topk=(1, 5)):
        super().__init__()
        assert isinstance(topk, (int, list, tuple))
        if isinstance(topk, int):
            topk = [topk]
        self.topk = topk

    def forward(self, x, label):
        return super().forward(x[0], label)
C
cuicheng01 已提交
241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 274 275 276 277 278 279 280 281 282 283 284 285 286 287 288 289 290 291 292 293 294 295 296 297 298 299 300 301 302 303 304 305 306 307 308


class MutiLabelMetric(object):
    def __init__(self):
        pass

    def _multi_hot_encode(self, logits, threshold=0.5):
        return binarize(logits, threshold=threshold)

    def __call__(self, output):
        output = F.sigmoid(output)
        preds = self._multi_hot_encode(logits=output.numpy(), threshold=0.5)
        return preds


class HammingDistance(MutiLabelMetric):
    """
    Soft metric based label for multilabel classification
    Returns:
        The smaller the return value is, the better model is.
    """

    def __init__(self):
        super().__init__()

    def __call__(self, output, target):
        preds = super().__call__(output)
        metric_dict = dict()
        metric_dict["HammingDistance"] = paddle.to_tensor(
            hamming_loss(target, preds))
        return metric_dict


class AccuracyScore(MutiLabelMetric):
    """
    Hard metric for multilabel classification
    Args:
        base: ["sample", "label"], default="sample"
            if "sample", return metric score based sample,
            if "label", return metric score based label.
    Returns:
        accuracy:
    """

    def __init__(self, base="label"):
        super().__init__()
        assert base in ["sample", "label"
                        ], 'must be one of ["sample", "label"]'
        self.base = base

    def __call__(self, output, target):
        preds = super().__call__(output)
        metric_dict = dict()
        if self.base == "sample":
            accuracy = accuracy_metric(target, preds)
        elif self.base == "label":
            mcm = multilabel_confusion_matrix(target, preds)
            tns = mcm[:, 0, 0]
            fns = mcm[:, 1, 0]
            tps = mcm[:, 1, 1]
            fps = mcm[:, 0, 1]
            accuracy = (sum(tps) + sum(tns)) / (
                sum(tps) + sum(tns) + sum(fns) + sum(fps))
            precision = sum(tps) / (sum(tps) + sum(fps))
            recall = sum(tps) / (sum(tps) + sum(fns))
            F1 = 2 * (accuracy * recall) / (accuracy + recall)
        metric_dict["AccuracyScore"] = paddle.to_tensor(accuracy)
        return metric_dict