utils.py

# 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 copy
import paddle
import paddle.nn as nn
import paddle.nn.functional as F

from ..bbox_utils import bbox_overlaps

__all__ = [
    '_get_clones', 'bbox_overlaps', 'bbox_cxcywh_to_xyxy',
    'bbox_xyxy_to_cxcywh', 'sigmoid_focal_loss'
]


def _get_clones(module, N):
    return nn.LayerList([copy.deepcopy(module) for _ in range(N)])


def bbox_cxcywh_to_xyxy(x):
    x_c, y_c, w, h = x.unbind(-1)
    b = [(x_c - 0.5 * w), (y_c - 0.5 * h), (x_c + 0.5 * w), (y_c + 0.5 * h)]
    return paddle.stack(b, axis=-1)


def bbox_xyxy_to_cxcywh(x):
    x0, y0, x1, y1 = x.unbind(-1)
    b = [(x0 + x1) / 2, (y0 + y1) / 2, (x1 - x0), (y1 - y0)]
    return paddle.stack(b, axis=-1)


def sigmoid_focal_loss(logit, label, normalizer=1.0, alpha=0.25, gamma=2.0):
    prob = F.sigmoid(logit)
    ce_loss = F.binary_cross_entropy_with_logits(logit, label, reduction="none")
    p_t = prob * label + (1 - prob) * (1 - label)
    loss = ce_loss * ((1 - p_t)**gamma)

    if alpha >= 0:
        alpha_t = alpha * label + (1 - alpha) * (1 - label)
        loss = alpha_t * loss
    return loss.mean(1).sum() / normalizer