Note

You are reading the documentation for MMSelfSup 0.x, which will soon be deprecated by the end of 2022. We recommend you upgrade to MMSelfSup 1.0.0rc versions to enjoy fruitful new features and better performance brought by OpenMMLab 2.0. Check out the changelog, code and documentation of MMSelfSup 1.0.0rc for more details.

Source code for mmselfsup.models.algorithms.deepcluster

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

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


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

    Implementation of `Deep Clustering for Unsupervised Learning
    of Visual Features <https://arxiv.org/abs/1807.05520>`_.
    The clustering operation is in `core/hooks/deepcluster_hook.py`.

    Args:
        backbone (dict): Config dict for module of backbone.
        with_sobel (bool): Whether to apply a Sobel filter on images.
            Defaults to True.
        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,
                 with_sobel=True,
                 neck=None,
                 head=None,
                 init_cfg=None):
        super(DeepCluster, self).__init__(init_cfg)
        self.with_sobel = with_sobel
        if with_sobel:
            self.sobel_layer = Sobel()
        self.backbone = build_backbone(backbone)
        if neck is not None:
            self.neck = build_neck(neck)
        assert head is not None
        self.head = build_head(head)

        # re-weight
        self.num_classes = self.head.num_classes
        self.loss_weight = torch.ones((self.num_classes, ),
                                      dtype=torch.float32)
        self.loss_weight /= self.loss_weight.sum()

    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.
        """
        if self.with_sobel:
            img = self.sobel_layer(img)
        x = self.backbone(img)
        return x

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

        Args:
            img (Tensor): Input images of shape (N, C, H, W).
                Typically these should be mean centered and std scaled.
            pseudo_label (Tensor): Label assignments.
            kwargs: Any keyword arguments to be used to forward.

        Returns:
            dict[str, Tensor]: A dictionary of loss components.
        """
        x = self.extract_feat(img)
        if self.with_neck:
            x = self.neck(x)
        outs = self.head(x)
        loss_inputs = (outs, pseudo_label)
        losses = self.head.loss(*loss_inputs)
        return losses

    def forward_test(self, img, **kwargs):
        """Forward computation during test.

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

        Returns:
            dict[str, Tensor]: A dictionary of output features.
        """
        x = self.extract_feat(img)  # tuple
        if self.with_neck:
            x = self.neck(x)
        outs = self.head(x)
        keys = [f'head{i}' for i in range(len(outs))]
        out_tensors = [out.cpu() for out in outs]  # NxC
        return dict(zip(keys, out_tensors))

    def set_reweight(self, labels, reweight_pow=0.5):
        """Loss re-weighting.

        Re-weighting the loss according to the number of samples in each class.

        Args:
            labels (numpy.ndarray): Label assignments.
            reweight_pow (float): The power of re-weighting. Defaults to 0.5.
        """
        histogram = np.bincount(
            labels, minlength=self.num_classes).astype(np.float32)
        inv_histogram = (1. / (histogram + 1e-10))**reweight_pow
        weight = inv_histogram / inv_histogram.sum()
        self.loss_weight.copy_(torch.from_numpy(weight))
        self.head.criterion = nn.CrossEntropyLoss(weight=self.loss_weight)