注意

您正在阅读 MMSelfSup 0.x 版本的文档，而 MMSelfSup 0.x 版本将会在 2022 年末 开始逐步停止维护。我们建议您及时升级到 MMSelfSup 1.0.0rc 版本，享受由 OpenMMLab 2.0 带来的更多新特性和更佳的性能表现。阅读 MMSelfSup 1.0.0rc 的 发版日志, 代码 和 文档 获取更多信息。

mmselfsup.models.algorithms.simsiam 源代码

# Copyright (c) OpenMMLab. All rights reserved.
import torch.nn as nn

from ..builder import ALGORITHMS, build_backbone, build_head, build_neck
from .base import BaseModel


@ALGORITHMS.register_module()
class SimSiam(BaseModel):
    """SimSiam.

    Implementation of `Exploring Simple Siamese Representation Learning
    <https://arxiv.org/abs/2011.10566>`_.
    The operation of fixing learning rate of predictor is in
    `core/hooks/simsiam_hook.py`.

    Args:
        backbone (dict): Config dict for module of backbone.
        neck (dict): Config dict for module of deep features to compact
            feature vectors. Defaults to None.
        head (dict): Config dict for module of loss functions.
            Defaults to None.
    """

    def __init__(self,
                 backbone,
                 neck=None,
                 head=None,
                 init_cfg=None,
                 **kwargs):
        super(SimSiam, self).__init__(init_cfg)
        assert neck is not None
        self.encoder = nn.Sequential(
            build_backbone(backbone), build_neck(neck))
        self.backbone = self.encoder[0]
        self.neck = self.encoder[1]
        assert head is not None
        self.head = build_head(head)

    def extract_feat(self, img):
        """Function to extract features from backbone.

        Args:
            img (Tensor): Input images of shape (N, C, H, W).
                Typically these should be mean centered and std scaled.

        Returns:
            tuple[Tensor]: backbone outputs.
        """
        x = self.backbone(img)
        return x

    def forward_train(self, img):
        """Forward computation during training.

        Args:
            img (list[Tensor]): A list of input images with shape
                (N, C, H, W). Typically these should be mean centered
                and std scaled.
        Returns:
            loss[str, Tensor]: A dictionary of loss components
        """
        assert isinstance(img, list)
        img_v1 = img[0]
        img_v2 = img[1]

        z1 = self.encoder(img_v1)[0]  # NxC
        z2 = self.encoder(img_v2)[0]  # NxC

        losses = 0.5 * (self.head(z1, z2)['loss'] + self.head(z2, z1)['loss'])
        return dict(loss=losses)