fcos.py 3.5 KB
Newer Older
F
Feng Ni 已提交
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19
# Copyright (c) 2020 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
F
Feng Ni 已提交
20
from ppdet.core.workspace import register, create
F
Feng Ni 已提交
21 22 23 24 25 26 27
from .meta_arch import BaseArch

__all__ = ['FCOS']


@register
class FCOS(BaseArch):
F
Feng Ni 已提交
28 29 30 31 32 33 34 35 36 37
    """
    FCOS network, see https://arxiv.org/abs/1904.01355

    Args:
        backbone (object): backbone instance
        neck (object): 'FPN' instance
        fcos_head (object): 'FCOSHead' instance
        post_process (object): 'FCOSPostProcess' instance
    """

F
Feng Ni 已提交
38
    __category__ = 'architecture'
F
Feng Ni 已提交
39
    __inject__ = ['fcos_post_process']
F
Feng Ni 已提交
40 41 42 43 44 45 46 47 48 49 50 51

    def __init__(self,
                 backbone,
                 neck,
                 fcos_head='FCOSHead',
                 fcos_post_process='FCOSPostProcess'):
        super(FCOS, self).__init__()
        self.backbone = backbone
        self.neck = neck
        self.fcos_head = fcos_head
        self.fcos_post_process = fcos_post_process

F
Feng Ni 已提交
52 53 54 55 56 57
    @classmethod
    def from_config(cls, cfg, *args, **kwargs):
        backbone = create(cfg['backbone'])

        kwargs = {'input_shape': backbone.out_shape}
        neck = create(cfg['neck'], **kwargs)
F
Feng Ni 已提交
58

F
Feng Ni 已提交
59 60
        kwargs = {'input_shape': neck.out_shape}
        fcos_head = create(cfg['fcos_head'], **kwargs)
F
Feng Ni 已提交
61

F
Feng Ni 已提交
62 63 64 65 66
        return {
            'backbone': backbone,
            'neck': neck,
            "fcos_head": fcos_head,
        }
F
Feng Ni 已提交
67

F
Feng Ni 已提交
68 69 70 71
    def _forward(self):
        body_feats = self.backbone(self.inputs)
        fpn_feats = self.neck(body_feats)
        fcos_head_outs = self.fcos_head(fpn_feats, self.training)
F
Feng Ni 已提交
72
        if not self.training:
F
Feng Ni 已提交
73 74 75 76 77
            scale_factor = self.inputs['scale_factor']
            bboxes = self.fcos_post_process(fcos_head_outs, scale_factor)
            return bboxes
        else:
            return fcos_head_outs
F
Feng Ni 已提交
78 79 80 81 82

    def get_loss(self, ):
        loss = {}
        tag_labels, tag_bboxes, tag_centerness = [], [], []
        for i in range(len(self.fcos_head.fpn_stride)):
F
Feng Ni 已提交
83
            # labels, reg_target, centerness
F
Feng Ni 已提交
84 85 86 87 88 89 90 91 92 93
            k_lbl = 'labels{}'.format(i)
            if k_lbl in self.inputs:
                tag_labels.append(self.inputs[k_lbl])
            k_box = 'reg_target{}'.format(i)
            if k_box in self.inputs:
                tag_bboxes.append(self.inputs[k_box])
            k_ctn = 'centerness{}'.format(i)
            if k_ctn in self.inputs:
                tag_centerness.append(self.inputs[k_ctn])

F
Feng Ni 已提交
94 95
        fcos_head_outs = self._forward()
        loss_fcos = self.fcos_head.get_loss(fcos_head_outs, tag_labels,
F
Feng Ni 已提交
96 97 98 99 100 101 102
                                            tag_bboxes, tag_centerness)
        loss.update(loss_fcos)
        total_loss = paddle.add_n(list(loss.values()))
        loss.update({'loss': total_loss})
        return loss

    def get_pred(self):
F
Feng Ni 已提交
103 104
        bbox_pred, bbox_num = self._forward()
        output = {'bbox': bbox_pred, 'bbox_num': bbox_num}
F
Feng Ni 已提交
105
        return output