Unsupervised domain adaptation transfers knowledge from a labeled source domain to an unlabeled target domain, which assumes that the source domain is labeled correctly, and both source and target data are available during the adaptive process. But collecting large-scale datasets with fully precise annotations is expensive and time-consuming. Besides, due to data privacy and security issues, the source data is often inaccessible during domain adaptation and only unlabeled target data is available. Therefore, considering both the source domain data with noisy labels and the unavailability of source domain data during domain adaptation, this paper proposes an Adaptive Neighbors and Uncertainty Estimation (ANUE) method for Source-Free Unsupervised Domain Adaptation with Noisy Labels (SF-UDA-NL). To the best of our knowledge, there has been no prior research conducted on this particular issue. Specifically, since source domain data contains noisy labels, we propose a reweighted small-loss with uncertainty estimation to filter reliable samples for updating the dual-branch network. To prevent noise knowledge from misleading domain adaptation, we adopt a contrastive learning framework and design an adaptive neighbors module to help target samples generate more reliable pseudo-labels. Then, we design a reweighted contrastive loss at both class level and instance level based on uncertainty estimation to further enhance the network's classification performance. We conduct extensive experiments on three widely used datasets for unsupervised domain adaptation in image classification, and the results demonstrate the effectiveness and robustness of our method.

Most methods of multi-person pose estimation (MPPE) treat the human detection and keypoint localization separately. They need additional supervision like instance bounding boxes, or complex hand-crafted processes like RoI cropping or grouping. In this article, we propose a novel one-stage MPPE method, named COPE, which unifies human detection and keypoint regression into an end-to-end learnable framework. To handle the challenges plague one-stage MPPE, i.e., instance overlapping and misalignment of local and global context, we design contrastive constraints at two levels of semantic granularity and feature sampling strategies. Based on a whole-process differentiable pipeline, COPE establishes a simple yet effective framework for MPPE without additional instance-level supervision and resource-intensive modules like transformer. Benefit from specially designed contrastive constraints and sampling strategies, COPE can better handle occluded scenes and correct keypoint localization errors. Extensive experiments demonstrate COPE’s superiority. It attains 71.3 AP and 18.0 FPS on COCO val2017, effectively balancing accuracy and speed. Particularly in crowded and occluded scenarios, COPE achieves state-of-the-art performance on CrowdPose and OCHuman, surpassing CID by 0.6 AP and 1.7 AP, respectively. Furthermore, COPE strongly improves generalization performance on the Human-Art benchmark, outperforming ED-Pose by 6.7 AP and ClickPose by 3.7 AP.

Just like many other topics in computer vision, image classification has achieved significant progress recently by using deep learning neural networks, especially the Convolutional Neural Networks (CNNs). Most of the existing works focused on classifying very clear natural images, evidenced by the widely used image databases, such as Caltech-256, PASCAL VOCs, and ImageNet. However, in many real applications, the acquired images may contain certain degradations that lead to various kinds of blurring, noise, and distortions. One important and interesting problem is the effect of such degradations to the performance of CNN-based image classification and whether degradation removal helps CNN-based image classification. More specifically, we wonder whether image classification performance drops with each kind of degradation, whether this drop can be avoided by including degraded images into training, and whether existing computer vision algorithms that attempt to remove such degradations can help improve the image classification performance. In this article, we empirically study those problems for nine kinds of degraded images-hazy images, motion-blurred images, fish-eye images, underwater images, low resolution images, salt-and-peppered images, images with white Gaussian noise, Gaussian-blurred images, and out-of-focus images. We expect this article can draw more interests from the community to study the classification of degraded images.

Although clear image classification and degraded image restoration have been studied a lot, regrettably, the degraded image classification problem has been largely overlooked. Degraded images usually have low classification performance due to the presence of blur, noise and other imperfections. Existing methods improve classification performance of degraded images by simple restoration, fine-tuning or data augment techniques. However, they ignore the useful information guided by clear images, which leads to the performance gap between degraded images and clear images. In this article, we find that the category distribution, feature distribution and visual attention of degraded images are usually inconsistent with that of clear images. Motivated by this observation, we propose an end-to-end Consistency Guided Network, named CG-Net, for degraded image classification. More specifically, we first propose Category Consistency Loss (CCL) to guide the model to learn category distribution that is more consistent with clear images. Second, we propose Semantic Consistency Loss (SCL) to enforce the model to learn more robust feature representation that is guided by clear images. Third, we propose Visual Attention Alignment Loss (VAAL), which can align the semantically informative regions between clear images and degraded images to improve the performance of degraded images. Besides, our method is more general and suitable for various kinds of degraded images. We conduct extensive experiments on various degraded images. The experimental results show that the proposed method significantly outperforms baselines, which demonstrate the effectiveness of our proposed method.

Degraded image semantic segmentation is of great importance in autonomous driving, highway navigation systems, and many other safety-related applications and it was not systematically studied before. In general, image degradations increase the difficulty of semantic segmentation, usually leading to decreased semantic segmentation accuracy. Therefore, performance on the underlying clean images can be treated as an upper bound of degraded image semantic segmentation. While the use of supervised deep learning has substantially improved the state of the art of semantic image segmentation, the gap between the feature distribution learned using the clean images and the feature distribution learned using the degraded images poses a major obstacle in improving the degraded image semantic segmentation performance. The conventional strategies for reducing the gap include: 1) Adding image-restoration based pre-processing modules; 2) Using both clean and the degraded images for training; 3) Fine-tuning the network pre-trained on the clean image. In this paper, we propose a novel Dense-Gram Network to more effectively reduce the gap than the conventional strategies and segment degraded images. Extensive experiments demonstrate that the proposed Dense-Gram Network yields stateof-the-art semantic segmentation performance on degraded images synthesized using PASCAL VOC 2012, SUNRGBD, CamVid, and CityScapes datasets.