test_yolov3_loss_op.py 8.5 KB
Newer Older
1 2 3 4 5 6 7 8 9 10 11 12 13 14
#   Copyright (c) 2018 PaddlePaddle Authors. All Rights Reserved.
#
# 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.

15 16
from __future__ import division

17 18
import unittest
import numpy as np
19 20
from scipy.special import logit
from scipy.special import expit
21 22
from op_test import OpTest

23 24
from paddle.fluid import core

D
dengkaipeng 已提交
25

26 27
def l1loss(x, y):
    return abs(x - y)
28 29


30
def sce(x, label):
31 32 33
    sigmoid_x = expit(x)
    term1 = label * np.log(sigmoid_x)
    term2 = (1.0 - label) * np.log(1.0 - sigmoid_x)
34
    return -term1 - term2
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
def sigmoid(x):
    return 1.0 / (1.0 + np.exp(-1.0 * x))


def batch_xywh_box_iou(box1, box2):
    b1_left = box1[:, :, 0] - box1[:, :, 2] / 2
    b1_right = box1[:, :, 0] + box1[:, :, 2] / 2
    b1_top = box1[:, :, 1] - box1[:, :, 3] / 2
    b1_bottom = box1[:, :, 1] + box1[:, :, 3] / 2

    b2_left = box2[:, :, 0] - box2[:, :, 2] / 2
    b2_right = box2[:, :, 0] + box2[:, :, 2] / 2
    b2_top = box2[:, :, 1] - box2[:, :, 3] / 2
    b2_bottom = box2[:, :, 1] + box2[:, :, 3] / 2

    left = np.maximum(b1_left[:, :, np.newaxis], b2_left[:, np.newaxis, :])
    right = np.minimum(b1_right[:, :, np.newaxis], b2_right[:, np.newaxis, :])
    top = np.maximum(b1_top[:, :, np.newaxis], b2_top[:, np.newaxis, :])
    bottom = np.minimum(b1_bottom[:, :, np.newaxis],
                        b2_bottom[:, np.newaxis, :])

    inter_w = np.clip(right - left, 0., 1.)
    inter_h = np.clip(bottom - top, 0., 1.)
    inter_area = inter_w * inter_h

    b1_area = (b1_right - b1_left) * (b1_bottom - b1_top)
    b2_area = (b2_right - b2_left) * (b2_bottom - b2_top)
    union = b1_area[:, :, np.newaxis] + b2_area[:, np.newaxis, :] - inter_area

    return inter_area / union


D
dengkaipeng 已提交
69
def YOLOv3Loss(x, gtbox, gtlabel, gtscore, attrs):
70 71 72 73 74 75 76 77 78
    n, c, h, w = x.shape
    b = gtbox.shape[1]
    anchors = attrs['anchors']
    an_num = len(anchors) // 2
    anchor_mask = attrs['anchor_mask']
    mask_num = len(anchor_mask)
    class_num = attrs["class_num"]
    ignore_thresh = attrs['ignore_thresh']
    downsample = attrs['downsample']
T
tink2123 已提交
79
    use_label_smooth = attrs['use_label_smooth']
80 81 82 83
    input_size = downsample * h
    x = x.reshape((n, mask_num, 5 + class_num, h, w)).transpose((0, 1, 3, 4, 2))
    loss = np.zeros((n)).astype('float32')

D
dengkaipeng 已提交
84 85 86
    label_pos = 1.0 - 1.0 / class_num if use_label_smooth else 1.0
    label_neg = 1.0 / class_num if use_label_smooth else 0.0

87 88 89 90 91 92
    pred_box = x[:, :, :, :, :4].copy()
    grid_x = np.tile(np.arange(w).reshape((1, w)), (h, 1))
    grid_y = np.tile(np.arange(h).reshape((h, 1)), (1, w))
    pred_box[:, :, :, :, 0] = (grid_x + sigmoid(pred_box[:, :, :, :, 0])) / w
    pred_box[:, :, :, :, 1] = (grid_y + sigmoid(pred_box[:, :, :, :, 1])) / h

D
dengkaipeng 已提交
93 94 95 96
    x[:, :, :, :, 5:] = np.where(x[:, :, :, :, 5:] < -0.5, x[:, :, :, :, 5:],
                                 np.ones_like(x[:, :, :, :, 5:]) * 1.0 /
                                 class_num)

97 98 99 100 101 102 103 104 105 106 107 108
    mask_anchors = []
    for m in anchor_mask:
        mask_anchors.append((anchors[2 * m], anchors[2 * m + 1]))
    anchors_s = np.array(
        [(an_w / input_size, an_h / input_size) for an_w, an_h in mask_anchors])
    anchor_w = anchors_s[:, 0:1].reshape((1, mask_num, 1, 1))
    anchor_h = anchors_s[:, 1:2].reshape((1, mask_num, 1, 1))
    pred_box[:, :, :, :, 2] = np.exp(pred_box[:, :, :, :, 2]) * anchor_w
    pred_box[:, :, :, :, 3] = np.exp(pred_box[:, :, :, :, 3]) * anchor_h

    pred_box = pred_box.reshape((n, -1, 4))
    pred_obj = x[:, :, :, :, 4].reshape((n, -1))
D
dengkaipeng 已提交
109
    objness = np.zeros(pred_box.shape[:2]).astype('float32')
110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126
    ious = batch_xywh_box_iou(pred_box, gtbox)
    ious_max = np.max(ious, axis=-1)
    objness = np.where(ious_max > ignore_thresh, -np.ones_like(objness),
                       objness)

    gtbox_shift = gtbox.copy()
    gtbox_shift[:, :, 0] = 0
    gtbox_shift[:, :, 1] = 0

    anchors = [(anchors[2 * i], anchors[2 * i + 1]) for i in range(0, an_num)]
    anchors_s = np.array(
        [(an_w / input_size, an_h / input_size) for an_w, an_h in anchors])
    anchor_boxes = np.concatenate(
        [np.zeros_like(anchors_s), anchors_s], axis=-1)
    anchor_boxes = np.tile(anchor_boxes[np.newaxis, :, :], (n, 1, 1))
    ious = batch_xywh_box_iou(gtbox_shift, anchor_boxes)
    iou_matches = np.argmax(ious, axis=-1)
127
    gt_matches = iou_matches.copy()
128 129 130
    for i in range(n):
        for j in range(b):
            if gtbox[i, j, 2:].sum() == 0:
131
                gt_matches[i, j] = -1
132 133
                continue
            if iou_matches[i, j] not in anchor_mask:
134
                gt_matches[i, j] = -1
135 136
                continue
            an_idx = anchor_mask.index(iou_matches[i, j])
137
            gt_matches[i, j] = an_idx
138 139 140 141 142 143 144 145 146 147 148 149 150
            gi = int(gtbox[i, j, 0] * w)
            gj = int(gtbox[i, j, 1] * h)

            tx = gtbox[i, j, 0] * w - gi
            ty = gtbox[i, j, 1] * w - gj
            tw = np.log(gtbox[i, j, 2] * input_size / mask_anchors[an_idx][0])
            th = np.log(gtbox[i, j, 3] * input_size / mask_anchors[an_idx][1])
            scale = 2.0 - gtbox[i, j, 2] * gtbox[i, j, 3]
            loss[i] += sce(x[i, an_idx, gj, gi, 0], tx) * scale
            loss[i] += sce(x[i, an_idx, gj, gi, 1], ty) * scale
            loss[i] += l1loss(x[i, an_idx, gj, gi, 2], tw) * scale
            loss[i] += l1loss(x[i, an_idx, gj, gi, 3], th) * scale

D
dengkaipeng 已提交
151
            objness[i, an_idx * h * w + gj * w + gi] = gtscore[i, j]
152 153

            for label_idx in range(class_num):
D
dengkaipeng 已提交
154 155
                loss[i] += sce(x[i, an_idx, gj, gi, 5 + label_idx], label_pos
                               if label_idx == gtlabel[i, j] else label_neg)
156 157 158 159 160

        for j in range(mask_num * h * w):
            if objness[i, j] >= 0:
                loss[i] += sce(pred_obj[i, j], objness[i, j])

D
dengkaipeng 已提交
161
    return (loss, objness.reshape((n, mask_num, h, w)).astype('float32'), \
162
            gt_matches.astype('int32'))
163 164


165 166 167 168
class TestYolov3LossOp(OpTest):
    def setUp(self):
        self.initTestCase()
        self.op_type = 'yolov3_loss'
169
        x = logit(np.random.uniform(0, 1, self.x_shape).astype('float32'))
170
        gtbox = np.random.random(size=self.gtbox_shape).astype('float32')
D
dengkaipeng 已提交
171
        gtlabel = np.random.randint(0, self.class_num, self.gtbox_shape[:2])
D
dengkaipeng 已提交
172
        gtscore = np.random.random(self.gtbox_shape[:2]).astype('float32')
D
dengkaipeng 已提交
173 174 175
        gtmask = np.random.randint(0, 2, self.gtbox_shape[:2])
        gtbox = gtbox * gtmask[:, :, np.newaxis]
        gtlabel = gtlabel * gtmask
176 177 178

        self.attrs = {
            "anchors": self.anchors,
179
            "anchor_mask": self.anchor_mask,
180 181
            "class_num": self.class_num,
            "ignore_thresh": self.ignore_thresh,
182
            "downsample": self.downsample,
183
            "use_label_smooth": self.use_label_smooth,
184 185
        }

D
dengkaipeng 已提交
186 187 188
        self.inputs = {
            'X': x,
            'GTBox': gtbox.astype('float32'),
D
dengkaipeng 已提交
189 190
            'GTLabel': gtlabel.astype('int32'),
            'GTScore': gtscore.astype('float32')
D
dengkaipeng 已提交
191
        }
D
dengkaipeng 已提交
192 193
        loss, objness, gt_matches = YOLOv3Loss(x, gtbox, gtlabel, gtscore,
                                               self.attrs)
194 195 196 197 198
        self.outputs = {
            'Loss': loss,
            'ObjectnessMask': objness,
            "GTMatchMask": gt_matches
        }
199 200

    def test_check_output(self):
201
        place = core.CPUPlace()
202
        self.check_output_with_place(place, atol=2e-3)
203

D
dengkaipeng 已提交
204 205 206 207 208
    def test_check_grad_ignore_gtbox(self):
        place = core.CPUPlace()
        self.check_grad_with_place(
            place, ['X'],
            'Loss',
D
dengkaipeng 已提交
209 210
            no_grad_set=set(["GTBox", "GTLabel", "GTScore"]),
            max_relative_error=0.2)
211 212

    def initTestCase(self):
213 214 215 216 217
        self.anchors = [
            10, 13, 16, 30, 33, 23, 30, 61, 62, 45, 59, 119, 116, 90, 156, 198,
            373, 326
        ]
        self.anchor_mask = [0, 1, 2]
D
dengkaipeng 已提交
218
        self.class_num = 10
219 220 221
        self.ignore_thresh = 0.7
        self.downsample = 32
        self.x_shape = (3, len(self.anchor_mask) * (5 + self.class_num), 5, 5)
D
dengkaipeng 已提交
222
        self.gtbox_shape = (3, 10, 4)
223 224 225
        self.use_label_smooth = True


D
dengkaipeng 已提交
226
class TestYolov3LossWithoutLabelSmooth(TestYolov3LossOp):
227
    def set_label_smooth(self):
D
dengkaipeng 已提交
228
        self.use_label_smooth = False
229 230 231 232


if __name__ == "__main__":
    unittest.main()