File size: 8,337 Bytes
7b7527a |
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 |
# Copyright (c) 2021 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.
from __future__ import absolute_import
from __future__ import division
from __future__ import print_function
import paddle
import paddle.nn as nn
import paddle.nn.functional as F
from ppdet.core.workspace import register
from ..bbox_utils import batch_iou_similarity
from .utils import (gather_topk_anchors, check_points_inside_bboxes,
compute_max_iou_anchor)
__all__ = ['TaskAlignedAssigner']
def is_close_gt(anchor, gt, stride_lst, max_dist=2.0, alpha=2.):
"""Calculate distance ratio of box1 and box2 in batch for larger stride
anchors dist/stride to promote the survive of large distance match
Args:
anchor (Tensor): box with the shape [L, 2]
gt (Tensor): box with the shape [N, M2, 4]
Return:
dist (Tensor): dist ratio between box1 and box2 with the shape [N, M1, M2]
"""
center1 = anchor.unsqueeze(0)
center2 = (gt[..., :2] + gt[..., -2:]) / 2.
center1 = center1.unsqueeze(1) # [N, M1, 2] -> [N, 1, M1, 2]
center2 = center2.unsqueeze(2) # [N, M2, 2] -> [N, M2, 1, 2]
stride = paddle.concat([
paddle.full([x], 32 / pow(2, idx)) for idx, x in enumerate(stride_lst)
]).unsqueeze(0).unsqueeze(0)
dist = paddle.linalg.norm(center1 - center2, p=2, axis=-1) / stride
dist_ratio = dist
dist_ratio[dist < max_dist] = 1.
dist_ratio[dist >= max_dist] = 0.
return dist_ratio
@register
class TaskAlignedAssigner(nn.Layer):
"""TOOD: Task-aligned One-stage Object Detection
"""
def __init__(self,
topk=13,
alpha=1.0,
beta=6.0,
eps=1e-9,
is_close_gt=False):
super(TaskAlignedAssigner, self).__init__()
self.topk = topk
self.alpha = alpha
self.beta = beta
self.eps = eps
self.is_close_gt = is_close_gt
@paddle.no_grad()
def forward(self,
pred_scores,
pred_bboxes,
anchor_points,
num_anchors_list,
gt_labels,
gt_bboxes,
pad_gt_mask,
bg_index,
gt_scores=None):
r"""This code is based on
https://github.com/fcjian/TOOD/blob/master/mmdet/core/bbox/assigners/task_aligned_assigner.py
The assignment is done in following steps
1. compute alignment metric between all bbox (bbox of all pyramid levels) and gt
2. select top-k bbox as candidates for each gt
3. limit the positive sample's center in gt (because the anchor-free detector
only can predict positive distance)
4. if an anchor box is assigned to multiple gts, the one with the
highest iou will be selected.
Args:
pred_scores (Tensor, float32): predicted class probability, shape(B, L, C)
pred_bboxes (Tensor, float32): predicted bounding boxes, shape(B, L, 4)
anchor_points (Tensor, float32): pre-defined anchors, shape(L, 2), "cxcy" format
num_anchors_list (List): num of anchors in each level, shape(L)
gt_labels (Tensor, int64|int32): Label of gt_bboxes, shape(B, n, 1)
gt_bboxes (Tensor, float32): Ground truth bboxes, shape(B, n, 4)
pad_gt_mask (Tensor, float32): 1 means bbox, 0 means no bbox, shape(B, n, 1)
bg_index (int): background index
gt_scores (Tensor|None, float32) Score of gt_bboxes, shape(B, n, 1)
Returns:
assigned_labels (Tensor): (B, L)
assigned_bboxes (Tensor): (B, L, 4)
assigned_scores (Tensor): (B, L, C)
"""
assert pred_scores.ndim == pred_bboxes.ndim
assert gt_labels.ndim == gt_bboxes.ndim and \
gt_bboxes.ndim == 3
batch_size, num_anchors, num_classes = pred_scores.shape
_, num_max_boxes, _ = gt_bboxes.shape
# negative batch
if num_max_boxes == 0:
assigned_labels = paddle.full(
[batch_size, num_anchors], bg_index, dtype='int32')
assigned_bboxes = paddle.zeros([batch_size, num_anchors, 4])
assigned_scores = paddle.zeros(
[batch_size, num_anchors, num_classes])
return assigned_labels, assigned_bboxes, assigned_scores
# compute iou between gt and pred bbox, [B, n, L]
ious = batch_iou_similarity(gt_bboxes, pred_bboxes)
# gather pred bboxes class score
pred_scores = pred_scores.transpose([0, 2, 1])
batch_ind = paddle.arange(
end=batch_size, dtype=gt_labels.dtype).unsqueeze(-1)
gt_labels_ind = paddle.stack(
[batch_ind.tile([1, num_max_boxes]), gt_labels.squeeze(-1)],
axis=-1)
bbox_cls_scores = paddle.gather_nd(pred_scores, gt_labels_ind)
# compute alignment metrics, [B, n, L]
alignment_metrics = bbox_cls_scores.pow(self.alpha) * ious.pow(
self.beta)
# check the positive sample's center in gt, [B, n, L]
if self.is_close_gt:
is_in_gts = is_close_gt(anchor_points, gt_bboxes, num_anchors_list)
else:
is_in_gts = check_points_inside_bboxes(anchor_points, gt_bboxes)
# select topk largest alignment metrics pred bbox as candidates
# for each gt, [B, n, L]
is_in_topk = gather_topk_anchors(
alignment_metrics * is_in_gts, self.topk, topk_mask=pad_gt_mask)
# select positive sample, [B, n, L]
mask_positive = is_in_topk * is_in_gts * pad_gt_mask
# if an anchor box is assigned to multiple gts,
# the one with the highest iou will be selected, [B, n, L]
mask_positive_sum = mask_positive.sum(axis=-2)
if mask_positive_sum.max() > 1:
mask_multiple_gts = (mask_positive_sum.unsqueeze(1) > 1).tile(
[1, num_max_boxes, 1])
is_max_iou = compute_max_iou_anchor(ious)
mask_positive = paddle.where(mask_multiple_gts, is_max_iou,
mask_positive)
mask_positive_sum = mask_positive.sum(axis=-2)
assigned_gt_index = mask_positive.argmax(axis=-2)
# assigned target
assigned_gt_index = assigned_gt_index + batch_ind * num_max_boxes
assigned_labels = paddle.gather(
gt_labels.flatten(), assigned_gt_index.flatten(), axis=0)
assigned_labels = assigned_labels.reshape([batch_size, num_anchors])
assigned_labels = paddle.where(
mask_positive_sum > 0, assigned_labels,
paddle.full_like(assigned_labels, bg_index))
assigned_bboxes = paddle.gather(
gt_bboxes.reshape([-1, 4]), assigned_gt_index.flatten(), axis=0)
assigned_bboxes = assigned_bboxes.reshape([batch_size, num_anchors, 4])
assigned_scores = F.one_hot(assigned_labels, num_classes + 1)
ind = list(range(num_classes + 1))
ind.remove(bg_index)
assigned_scores = paddle.index_select(
assigned_scores, paddle.to_tensor(ind), axis=-1)
# rescale alignment metrics
alignment_metrics *= mask_positive
max_metrics_per_instance = alignment_metrics.max(axis=-1, keepdim=True)
max_ious_per_instance = (ious * mask_positive).max(axis=-1,
keepdim=True)
alignment_metrics = alignment_metrics / (
max_metrics_per_instance + self.eps) * max_ious_per_instance
alignment_metrics = alignment_metrics.max(-2).unsqueeze(-1)
assigned_scores = assigned_scores * alignment_metrics
return assigned_labels, assigned_bboxes, assigned_scores, mask_positive
|