Test Time Adaptation for Blind Image Quality Assessment
测试时自适应用于无参考图像质量评估

One of the first pieces of work on TTA [35] introduces a joint training framework using a loss for the main task and a self-supervised auxiliary task loss. The choice of a relevant auxiliary task is a challenging part of the design. Sun et al. [35] use rotation prediction as a pre-text task for image classification applications. Researchers also explore simpler tasks that are highly correlated with the main task, such as feature alignment with the batch and a simple contrastive loss framework [24] for adaptation. However, these methods require the source data to train the source model all over again to enable TTA.
关于 TTA 的首批工作之一[35]介绍了一个联合训练框架，使用主任务的损失和自监督辅助任务损失。选择相关的辅助任务是设计中的一个挑战部分。Sun 等人[35]在图像分类应用中使用旋转预测作为前文本任务。研究人员还探索了与主任务高度相关的更简单任务，例如与批次的特征对齐和一个简单的对比损失框架[24]用于适应。然而，这些方法需要源数据重新训练源模型，才能实现 TTA。

TENT [37] studies source-free TTA, where entropy minimization-based adaptation even outperforms some methods that use the source data. Moreover, only the batch normalization parameters of the model are adapted in TENT. There are several works on batch norm statistics adaptation [29, 32] to improve the robustness of the model at test time. SHOT [20] presents a clustering-based pseudo-labeling method to align features from the target domain to the source domain using an information maximization loss. While a plethora of methods exists as above, the auxiliary tasks in these methods are not relevant for IQA.
TENT [37] 研究了无源 TTA，其中基于熵最小化的适应性甚至优于一些使用源数据的方法。此外，在 TENT 中，只有模型的批归一化参数进行了适应。有几项关于批归一化统计量适应的工作 [29, 32] 以提高模型在测试时的鲁棒性。SHOT [20] 提出了一种基于聚类的伪标记方法，通过信息最大化损失来对齐目标域与源域的特征。虽然存在许多如上所述的方法，但这些方法中的辅助任务与 IQA 不相关。

Most classical NR IQA algorithms are mainly based on natural scene statistics (NSS) [26, 42, 23]. In [27], the naturalness of the distorted image in the wavelet domain is modeled based on NSS. Saad et al. [31] also design the NSS model in the discrete cosine transform (DCT) domain. CORNIA [39], and HOSA [38] are among the earliest codebook learning based methods to predict quality. With the emergence of deep learning and the availability of large subject-rated IQA databases, various general-purpose NR IQA methods have been designed based on convolutional neural network (CNN) architectures [40, 22, 2, 25]. Some of these methods require end-to-end training [2, 25] of deep neural networks (DNN), while others [43, 34, 9, 15] are based on updating pre-trained models with some modifications.
大多数经典的无参考图像质量评估（NR IQA）算法主要基于自然场景统计（NSS）[26, 42, 23]。在[27]中，基于 NSS 在小波域中对失真图像的自然性进行了建模。Saad 等人[31]也在离散余弦变换（DCT）域设计了 NSS 模型。CORNIA [39]和 HOSA [38]是基于词典学习的最早的方法之一，用于预测质量。随着深度学习的兴起以及大规模主观评价 IQA 数据库的出现，已经设计了多种基于卷积神经网络（CNN）架构的一般用途的 NR IQA 方法[40, 22, 2, 25]。其中一些方法需要对深度神经网络（DNN）进行端到端训练[2, 25]，而其他方法[43, 34, 9, 15]则基于对预训练模型进行一些修改的更新。

More recently, Zhang et al. [43] propose a method to train bilinear DNNs to simultaneously model both authentic and synthetic distortions. Su et al. [34] design a self-adaptive hyper network to provide weights for the quality prediction module parameters and handle various types of distortions and content in the images. Several methods [9, 15] explore transformer-based architectures along with a CNN to capture dependencies between local and global features. MetaIQA [44] proposes meta-learning on synthetic distortions by using a shared quality prior knowledge model to adapt to any kind of distortion.
最近，Zhang 等人[43]提出了一种训练双线性 DNN 的方法，以同时模拟真实和合成失真。Su 等人[34]设计了一种自适应超网络，为质量预测模块参数提供权重，并处理图像中的各种失真和内容。几种方法[9, 15]探讨了基于 Transformer 的架构与 CNN 结合，以捕捉局部和全局特征之间的依赖关系。MetaIQA[44]通过使用共享质量先验知识模型进行元学习，在合成失真方面提出了一种可以适应任何失真的方法。

All the existing methods assume that the train and test data come from the same distribution. If there is a distribution shift across different databases, we need adaptation of the pre-trained model to learn about the target distributional information.
所有现有的方法都默认训练数据和测试数据来自同一分布。如果在不同数据库之间存在分布偏差，我们需要对预训练模型进行适应，以了解目标分布的信息。

Let the model trained on the source data be $f_{\theta}$ where $\theta=(\theta_{e},\theta_{c})$ corresponds to the parameters of the network, and $\theta_{e}$ and $\theta_{c}$ correspond to the parameters of the feature extractor and regression layers. Thus, for an input image $x$, we denote $f_{\theta}(x)=f_{\theta_{c}}(f_{\theta_{e}}(x))$. Since our goal is to learn the distribution shift between train and test data, we only update the parameters of the feature extractor, $\theta_{e}$, to align the features between the train and test distributions in a lower dimensional space.
让在源数据上训练的模型为 $f_{\theta}$ ，其中 $\theta=(\theta_{e},\theta_{c})$ 对应网络的参数， $\theta_{e}$ 和 $\theta_{c}$ 对应特征提取器和回归层的参数。因此，对于输入图像 $x$ ，我们表示为 $f_{\theta}(x)=f_{\theta_{c}}(f_{\theta_{e}}(x))$ 。由于我们的目标是学习训练数据和测试数据之间的分布偏移，我们只更新特征提取器的参数 $\theta_{e}$ ，以在低维空间中对齐训练和测试分布之间的特征。

The key challenge in TTA is the choice of a self-supervised auxiliary task that is highly correlated with the main task of IQA. However, relying too much on the auxiliary task can affect the performance of the main task. To prevent loss of learnt information induced by the auxiliary task [4], we first project the features to a lower dimensional space using a non-linear projection head $f_{\theta_{s}}$, parameterized by $\theta_{s}$. The auxiliary task drives the adaptation of the feature extractor through the projection head. Thus our model now has three parts parametrized by $(\theta_{e},\theta_{s},\theta_{c})$ to resemble a Y-shape as shown in Figure 1.
TTA 的关键挑战在于选择一个与 IQA 主要任务高度相关的自监督辅助任务。然而，过度依赖辅助任务可能会影响主要任务的表现。为了防止因辅助任务而导致的学习信息丢失[4]，我们首先使用一个非线性投射头 $f_{\theta_{s}}$ 来将特征投射到低维空间，该投射头由 $\theta_{s}$ 参数化。辅助任务通过投射头驱动特征提取器的适应。因此，我们的模型现在有三部分通过 $(\theta_{e},\theta_{s},\theta_{c})$ 参数化，类似于如图 1 所示的 Y 形。

When a batch of test instances $D$ arrives, we extract features from the feature extractor, followed by the projection head, and update the set of parameters ($\theta_{e},\theta_{s})$ by optimizing a self-supervised objective function $\mathcal{L}_{s}(D)$. Updating all the model parameters of the feature extractor $\theta_{e}$ can cause the model to diverge too much from training, and the performance can drop drastically. Inspired by prior work on test time entropy minimization [37] and improving robustness for test data [32], we only update the linear and lower-dimensional feature modulation parameters. In a neural network, normalization layers satisfy these properties. So we only adapt the batch normalization layers by updating the affine parameters to mitigate distributional shifts. Thus we update these parameters by optimizing the auxiliary task loss given by
当一批测试实例 $D$ 到达时，我们从特征提取器中提取特征，然后通过投影头，并通过优化一个自监督目标函数 $\mathcal{L}_{s}(D)$ 来更新参数集 $\theta_{e},\theta_{s})$ 。更新特征提取器 $\theta_{e}$ 的所有模型参数可能会导致模型与训练偏离过多，表现急剧下降。受到测试时间熵最小化[37]和提高测试数据鲁棒性[32]方面的早期工作的启发，我们仅更新线性和低维特征调制参数。在神经网络中，归一化层满足这些特性。因此我们仅通过更新仿射参数来调整批量归一化层，以减轻分布变化。因此，我们通过优化给出的辅助任务损失更新这些参数。

After optimizing the above loss function, we use the updated feature extractor parameters $\theta_{e}^{*}$ and pre-trained quality regressor parameters $\theta_{c}$ to predict the quality for a batch of test data. Since the distribution across batches can vary significantly, we ignore the earlier updated model and start the TTA based on source model weights for new test instants. Thus the updated target model always only depends on the incoming test data.
在优化上述损失函数后，我们使用更新的特征提取器参数 $\theta_{e}^{*}$ 和预训练的质量回归器参数 $\theta_{c}$ 来预测一批测试数据的质量。由于各批次之间的分布可能有显著差异，我们忽略之前更新的模型，并基于源模型权重开始新的测试时刻的 TTA。因此，更新后的目标模型始终仅依赖于输入的测试数据。

Our goal is to carefully choose self-supervised auxiliary tasks that capture quality-aware distributional information to adapt the feature encoder. We formulate two novel and complementary self-supervised learning techniques which help to learn the distribution shift between train and test data. These self-supervised objective functions are - 1) Group Contrastive Loss and 2) Rank Loss. While the GC loss works well when there is a reasonable separation of quality among the samples in a batch, the rank loss works better even when the quality of the batch samples is similar. On the other hand, the rank loss is meaningful only when the quality of the input image is not extremely low, while the GC loss is independent of the quality of a given image. Thus, a combination of the two losses renders our TTA extremely effective across various scenarios.
我们的目标是精心选择自监督辅助任务，以捕捉质量感知的分布信息来适应特征编码器。我们提出了两种新颖且互补的自监督学习技术，帮助学习训练和测试数据之间的分布偏差。这些自监督目标函数是：1）组对比损失和 2）排序损失。虽然当批次样本中质量有合理分离时，组对比损失效果较好，但即使批次样本质量相似，排序损失也能表现得更好。另一方面，排序损失只有在输入图像质量不是极低时才有意义，而组对比损失则与给定图像的质量无关。因此，两种损失的结合使我们的 TTA 在各种情况下都极其有效。

We evaluate TTA on four popular IQA databases using four different quality-aware models. In particular, we consider state-of-the-art deep IQA models such as TReS [9], MUSIQ [15] , HyperIQA [34], and MetaIQA [44]. These models contain a ResNet backbone with batch normalization layers, which we model as the feature extractor $f_{\theta_{e}}$ and only update its batch normalization parameters. We model the rest of the network as the quality regressor part, $f_{\theta_{c}}$. TReS and MUSIQ use transformers as a part of their architecture, which we include as a part of the quality regressor in all our main experiments. We also explore the adaptation of transformers by including them as part of the feature extractor and updating their layer normalization statistics in the supplementary material.
我们在四个流行的 IQA 数据库上使用四种不同的质量感知模型评估 TTA。特别是，我们考虑了最先进的深度 IQA 模型，例如 TReS [9]、MUSIQ [15]、HyperIQA [34]和 MetaIQA [44]。这些模型包含一个带有批归一化层的 ResNet 骨干，我们将其建模为特征提取器 $f_{\theta_{e}}$ ，并且仅更新其批归一化参数。我们将网络的其余部分建模为质量回归器部分 $f_{\theta_{c}}$ 。TReS 和 MUSIQ 在其架构中使用了 Transformer，我们在所有主要实验中将其作为质量回归器的一部分。我们还在补充材料中探索了通过将 Transformer 作为特征提取器的一部分并更新其层归一化统计来适应 Transformer。

We project the features extracted from the last layer of ResNet through a 256-dimensional fully connected (FC) layer corresponding to the self-supervised projection head $f_{\theta_{s}}$. These lower dimensional features are used to adapt the model at test time. Three IQA models, TReS, MUSIQ, and HyperIQA are trained on the LIVEFB database [41] containing 39,810 images. MetaIQA is trained on two synthetically distorted databases, TID2013 [30] and KADID-10k [21].
我们通过一个 256 维的全连接（FC）层将从 ResNet 最后一层提取的特征投影到对应的自监督投影头 $f_{\theta_{s}}$ 。这些低维特征用于在测试时调整模型。三个 IQA 模型，TReS、MUSIQ 和 HyperIQA 在 LIVEFB 数据库[41]上训练，该数据库包含 39,810 张图像。MetaIQA 则在两个合成的失真数据库 TID2013[30]和 KADID-10k[21]上进行训练。

We choose challenging databases to evaluate the generalization capability of our TTA-IQA method. The test datasets are described as follows:
我们选择了具有挑战性的数据集来评估我们的 TTA-IQA 方法的泛化能力。测试数据集描述如下：

KonIQ-10k [11] is a popular in-the-wild authentically distorted database consisting of 10,073 quality-scored images.
KonIQ-10k [11] 是一个流行的真实环境中自然失真的数据库，包含 10,073 张质量评分的图像。

PIPAL [13] is a large-scale IQA database for evaluating perceptual image restoration. It contains 29k distorted images, including 19 different GAN-based algorithms.
PIPAL [13] 是一个评估感知图像修复的大规模 IQA 数据库。它包含 29,000 张失真图像，包括 19 种不同的基于 GAN 的算法。

CID2013 [36] consists of 480 images in six image sets captured by 79 imaging devices.
CID2013 [36] 包含六组图像，共 480 张图像，由 79 台成像设备拍摄。

LIVE-IQA [33] consists of 779 synthetically distorted images from 29 reference images.
LIVE-IQA [33] 包含来自 29 张参考图像的 779 张合成失真图像。

We evaluate our results using Spearman’s rank-order correlation coefficient (SROCC) and Pearson’s linear correlation coefficient (PLCC).
我们使用斯皮尔曼等级相关系数（SROCC）和皮尔逊线性相关系数（PLCC）来评估我们的结果。

We implement our setup in PyTorch and conduct all the experiments with an 16 GB NVIDIA RTX A4000 GPU. During TTA, we randomly select a patch of size $224\times 224$ from the input image and apply quality preserving augmentations such as horizontal flip and vertical flip before passing it through the network for TTA. We use the ADAM [16] optimizer with a learning rate of 0.001 and set the number of iteration as 3. All the experiments were run on five different seeds using a batch size of 8, and the final results were obtained by averaging.
我们在 PyTorch 中实现了我们的设置，并使用一台 16GB NVIDIA RTX A4000 GPU 进行所有实验。在 TTA 期间，我们从输入图像中随机选择一个大小为 $224\times 224$ 的补丁，并在通过网络进行 TTA 之前应用质量保持增强，比如水平翻转和垂直翻转。我们使用学习率为 0.001 的 ADAM [16]优化器，并将迭代次数设定为 3。所有实验都在五个不同的种子上运行，使用的批量大小为 8，最终结果通过平均获得。

With regard to the self-supervised auxiliary tasks, we use a combination of rank loss and GC loss as our objective function using $\lambda=1$ given in Equation (7). We choose $p=0.25$ to determine the groups. We calculate the GC loss using $\tau=1$. For the rank loss, the test image is synthetically blurred using a Gaussian blur filter of size $5\times 5$ with two sets of standard deviations $\sigma$, for the Gaussian blur kernel. We keep $\sigma\in[40,80]$ for highly distorted images and $\sigma\in[1,20]$ for less blurred images. For compression distortions, we specify the quality factor in $[80,95]$ for lower compression and $[30,60]$ for higher compression rates. Similarly, we add zero mean Gaussian white noise to the test image with variance in $[0.05,0.1]$ for higher distortion and in $[0.005,0.01]$ for lower distortion.
关于自监督辅助任务，我们使用排名损失和 GC 损失的组合作为我们的目标函数，使用等式（7）中给出的 $\lambda=1$ 。我们选择 $p=0.25$ 来确定分组。我们使用 $\tau=1$ 计算 GC 损失。对于排名损失，测试图像通过大小为 $5\times 5$ 的高斯模糊滤波器进行合成模糊，使用两组标准差 $\sigma$ ，作为高斯模糊内核。我们保持 $\sigma\in[40,80]$ 用于高度失真图像， $\sigma\in[1,20]$ 用于较少模糊的图像。对于压缩失真，我们在 $[80,95]$ 中指定较低压缩的质量因数，以及在 $[30,60]$ 中指定较高压缩率。同样地，我们对测试图像添加零均值的高斯白噪声，其中 $[0.05,0.1]$ 表示较高失真， $[0.005,0.01]$ 表示较低失真。

Table 1 shows the performance of our method on all four test datasets using all four quality-aware models. In addition to the source model, we also compare with rotation prediction [35] as the auxiliary task during TTA. A comparison with such a task helps understand the role of quality-aware losses for the TTA of IQA models.
表 1 显示了我们的方法在所有四个测试数据集上，使用所有四个质量感知模型的性能。除了源模型之外，我们还将旋转预测[35]作为辅助任务进行比较。与这样的任务进行比较有助于理解质量感知损失在 IQATTA 中的作用。

We observe that TTA-IQA using the combination of rank and GC loss outperforms the source models in all the cases. Note that the PIPAL dataset has a huge distribution shift from the authentically distorted LIVEFB dataset on which three of the models were trained. While most source models perform very poorly on PIPAL, we achieve 10%-20% improvement over the source models. On the KonIQ-10k dataset, TTA-IQA gives around 1%-10% improvement over the source model. As both KonIQ-10k and LIVEFB are authentically distorted datasets, the distribution shift is reasonably small, leading to smaller improvements using adaptation. CID2013 is also an authentically distorted dataset and gives similar improvements of around 2%-13% over the source models. Our experiments on the LIVE-IQA database provide a significantly greater improvement of 10%-33% owing to the shift from authentic to synthetic distortions for three models.
我们观察到，结合排名和 GC 损失的 TTA-IQA 在所有情况下都优于源模型。需要注意的是，PIPAL 数据集与三种模型训练的真实失真 LIVEFB 数据集之间存在巨大分布差异。尽管大多数源模型在 PIPAL 上表现很差，我们相比源模型取得了 10%-20%的提升。在 KonIQ-10k 数据集上，TTA-IQA 相较于源模型提高了约 1%-10%。由于 KonIQ-10k 和 LIVEFB 都是真实失真数据集，分布差异相对较小，因此通过适应性得到的提升也较小。CID2013 也是一个真实失真数据集，相比源模型获得了约 2%-13%的相似提升。我们在 LIVE-IQA 数据库上的实验显示，由于三个模型从真实失真转向合成失真，改进显著提高了 10%-33%。

We also observe from Table 1 that our approach outperforms the rotation prediction task in most cases. Other TTA ideas, such as TENT [37] and training using masked autoencoders [7] are not applicable for the IQA task. In particular, the notion of entropy in regression tasks is not clear. On the other hand, masked reconstructions tend to lead to quality degradation. Thus, we do not compare with them.
我们还从表 1 中观察到，我们的方法在大多数情况下优于旋转预测任务。其他测试时间适应（TTA）方法，如 TENT [37]和使用掩码自动编码器训练[7]不适用于图像质量评估（IQA）任务。特别是在回归任务中熵的概念不明确。另一方面，掩码重建往往会导致质量下降。因此，我们没有与它们进行比较。

Need for Rank Loss and GC Loss for TTA. We perform an ablation study with respect to the two losses in Table 2. We see from the results that the rank loss and the GC loss individually always improve on the source models. Further, the combination of the rank and GC loss provides an even better performance over the individual losses in most evaluation scenarios. Even in scenarios, where the combination loss achieves the second-best performance, its performance is very close to the best. For the rest of the experiments in this section, we present results on all four datasets using the TReS method alone.
需要排名损失和 GC 损失用于 TTA。我们进行了一项关于表 2 中两个损失的消融研究。我们从结果中看到，排名损失和 GC 损失在独立应用时总能改善源模型的表现。此外，排名损失和 GC 损失的结合在大多数评估场景中表现出更好的性能。即便在组合损失表现为第二好的场景中，其表现也接近最佳。在本节余下的实验中，我们仅使用 TReS 方法展示所有四个数据集的结果。

We now discuss scenarios where the GC loss is particularly more useful than the rank loss. If the test image is extremely distorted, none of the three kinds of distortion types can create much difference in the perceptual quality of the two degraded versions. To understand this further, we consider highly distorted images (corresponding to a differential mean opinion score (DMOS) greater than 70) of different distortion types in the LIVE-IQA dataset. We apply TTA only for these images using different losses. The SROCC performance for these images in Figure 5 reveal that the rank loss is not very effective. However, the GC loss works well in such scenarios.
我们现在讨论 GC 损失比排名损失更有用的情况。如果测试图像极度失真，任何三种失真类型都无法在两个降级版本的感知质量中产生太大差异。为了进一步理解这一点，我们考虑 LIVE-IQA 数据集中不同失真类型的高度失真图像（对应于差异平均意见得分 (DMOS) 大于 70）。我们仅对这些图像应用 TTA 并使用不同的损失。图 5 中这些图像的 SROCC 表现表明排名损失效果不佳。然而，GC 损失在这样的情况下表现良好。

Conversely, we also discuss scenarios where the rank loss is more useful than the GC loss. To illustrate this idea, we select multiple images having similar quality with DMOS in $[28,44]$ from the LIVE-IQA dataset. We adapt the source model for this set of images using both the GC and rank losses. From Figure 4, we observe that the rank loss leads to much better adaptation in terms of the resulting model correlating with the ground truth scores. Intuitively, when the input images have a similar quality, the pseudo-labels given by the source model may be noisy, leading to an inaccurate grouping of the images into the lower and higher quality groups. This can adversely affect the GC loss. So the rank loss is more effective than the GC loss for test batches with similar quality images.
相反，我们也讨论了在某些情况下，秩损失比 GC 损失更有用。为了说明这个观点，我们从 LIVE-IQA 数据集中选择了多个质量相似且 DMOS 为 $[28,44]$ 的图像。我们使用 GC 损失和秩损失对这组图像进行源模型的适应。从图 4 中可以观察到，秩损失在模型与真实评分结果的相关性方面表现出更好的适应性。直观上，当输入图像具有相似质量时，由源模型给出的伪标签可能会很噪声，导致将图像不准确地分组到低质量和高质量组中。这可能会对 GC 损失产生不利影响。因此，对于具有类似质量图像的测试批次，秩损失比 GC 损失更为有效。

Effect of Selecting Multiple Distortion Types in the Rank Loss. We experiment with the impact of different kinds of distortion types while incorporating the rank loss. In particular, we consider three scenarios: one where only one of the distortion types is used in the rank loss, another where all the three types are used to obtain three loss terms which are summed together to update the model, and a third, where only one of the distortion types is chosen based on the discussion in Section 3.2.2. We observe from Table 3 that choosing the distortion type with the maximal difference in quality gives the best performance for the rank loss. This experiment validates our hypothesis that only a rank loss that can discriminate between the degraded image pairs is useful for TTA.
选择多种失真类型对排序损失的影响。我们通过结合排序损失，实验了不同类型失真对其影响。特别是，我们考虑三个场景：一个场景是仅使用一种失真类型进行排序损失，另一个场景是使用三种失真类型来获得三个损失项并相加以更新模型，第三个场景则根据 3.2.2 节中的讨论选择一种失真类型。从表 3 可以观察到，选择质量差异最大的失真类型能为排序损失带来最佳性能。该实验验证了我们的假设，即只有能区分降质图像对的排序损失才对 TTA 有用。

Selection of Group Size by Varying $p$. We explore different values of $p$ for constructing the GC loss. For a batch size of 8, possible choices of the group size are $2,3,4$ corresponding to $p=0.25,0.375,0.5$ respectively. From Table 4, we observe that the performances are roughly similar for different values of $p$ with a slightly superior performance for $p=0.25$. We note that as $p$ increases, the groups are larger and probably less discriminative in terms of quality, leading to a slightly poorer performance.
通过改变 $p$ 选择组大小。我们探索不同的 $p$ 值来构建 GC 损失。对于批量大小为 8 的情况，组大小的可能选择对应于分别为 $2,3,4$ 的 $p=0.25,0.375,0.5$ 。从表 4 可以看出，对于不同的 $p$ 值，性能大致相似，但 $p=0.25$ 略占优势。我们注意到随着 $p$ 增加，组变得更大，在质量方面可能不那么具有区分性，导致略差的性能。

Selection of Number of Groups for GC Loss. In our formulation, we define the GC loss only between two contrastive groups. In principle, we can extend our idea of GC loss to multiple groups. In particular, we can cluster the images in a batch into multiple groups and apply the GC loss between every pair of groups and sum the loss terms. Thus, images coming from the same group act as positive pairs, and images from two different groups are considered as negative pairs. From Table 5, we observe that the number of groups $G$ does not impact the performance much.
选择 GC 损失的组数。在我们的公式中，我们仅在两个对比组之间定义 GC 损失。原则上，我们可以将 GC 损失的概念扩展到多个组。具体来说，我们可以将一个批次中的图像聚类成多个组，并在每对组之间应用 GC 损失并求和损失项。因此，来自同一组的图像充当正对图像，而来自两个不同组的图像则视为负对图像。根据表 5，我们观察到组数 $G$ 对性能影响不大。

Choice of Number of Iterations for Learning Auxiliary Task. We also examine the effect of the number of iterations of parameter updates during TTA. Figure 6 shows that the performance on CID2013 dataset is fairly robust until 4 iterations. Beyond that, the model overfits the auxiliary task and leads to poor performance at test time.
学习辅助任务的迭代次数选择。我们还检查了 TTA 期间参数更新迭代次数的影响。图 6 显示，CID2013 数据集上的性能在进行 4 次迭代之前都相当稳健。超过这个次数后，模型则会对辅助任务产生过拟合，从而导致测试时性能较差。