Unsupervised anomaly detection has emerged as a powerful technique for identifying abnormal patterns in images without relying on pre-labeled defective samples. Many unsupervised methods use pre-trained feature extractors from large datasets, with knowledge distillation between teacher and student models being a leading technique. However, due to the similar structures of teacher and student, these methods face challenges like excessive specialization and inadequate generalization, reducing detection performance. In this paper, we introduce a Co-Progression Knowledge Distillation (CPKD) framework, enabling bidirectional learning between teacher and student models. This innovative framework enables concurrent evolution of both models, fostering mutual improvement and enhanced adaptability. To maintain system stability and prevent overspecialization, we introduce a knowledge prototype as a regulatory mechanism for the teacher's learning process. Our method effectively addresses key challenges in anomaly detection, including insufficient learning and overadaptation, by striking a balance between acquiring new knowledge and preserving core competencies. We demonstrate significant improvements in detection accuracy, achieving SOTA performance on the MVTec dataset.

Anomaly detection is a crucial aspect in ensuring the reliability of industrial products in smart manufacturing. Despite employing unsupervised anomaly detection methods to mitigate the challenges associated with obtaining defect data, there remains the challenge of adapting anomaly detectors to drift within normal data distributions, especially when there are significant intra-class variations in normal samples. Consequently, zero-shot anomaly detection techniques, which do not require training data and thus avoid interference from training data, have emerged as a new research direction. Meanwhile, the zero-shot segmentation capability demonstrated by the Segment Anything Mode (SAM) visual foundation model has further fueled our research enthusiasm. In this paper, we focus on texture images and propose a no-external prompt SAM-based method for zero-shot texture anomaly detection, called Image Internal Information Mining SAM (IIIM-SAM). We utilize the Image Internal Information Mining Prompter to mine relationships between internal regions of the image, identifying specific background and potential defect points. The two types of prompt points are fed into the SAM model in different stages, imparting the SAM with incorporating category information segmentation capability. Unlike current methods, our zero-shot texture anomaly detection method requires no external prompts or expert knowledge, and also avoids the uncertainty introduced by text prompts necessary for CLIP-based approaches. Our experimental results demonstrate the superiority of our zero-shot texture anomaly detection method compared to other approaches. Specifically, in the texture subset of MVTecAD/KolektorSDD2, our IIIM-SAM achieves 99.2%/93.6% image-level AUROC and 98.6%/92.3% pixel-level AUROC.

In autonomous driving, the LiDAR-Camera system plays a crucial role in a vehicle’s perception of 3D environments. To effectively fuse information from both camera and LiDAR, extrinsic calibration is indispensable. Recently, some researchers have proposed deep learning-based methods that utilize convolutional networks to automatically extract features from LiDAR depth images and RGB images for calibration. However, these features do not sufficiently interact during feature matching, which limits the calibration accuracy. To this end, we introduce a novel extrinsic calibration network (HIFM-Net) in this paper. It establishes a comprehensive connection between camera and LiDAR features by calculating a globally-aware map-to-map cost volume and hierachical point-to-map cost volumes. The former is used to regress large extrinsic offsets. The latter is employed to iteratively fine-tune extrinsic parameters, while the rigidity of LiDAR points is considered in each iteration to enhance regression robustness. Extensive experiments on the KITTI-odometry dataset demonstrate the superior performance of our HIFMNet compared to other state-of-the-art learning-based methods.

As an emerging vision foundation model, Segment Anything Model (SAM) has been successfully applied to Video Object Segmentation (VOS). However, previous methods rely on first frame’s mask or manual interaction, which belong to semi-supervised or interactive video object segmentation. The potential of SAM in Unsupervised Video Object Segmentation (UVOS) remains to be explored. In this work, we propose a two-stage training framework to explore how SAM helps UVOS. In Stage 1, utilizing SAM powerful feature extraction ability, we only train a lightweight decoder with feature aggregation module. This stage employs relaxed flow reconstruction as an unsupervised proxy task for object discovery. In Stage 2, based on SAM promptability, we design a refinement process that involves rough correction based on in-clip temporal consistency and heuristic refinement guided by Prompt-closure, to generate refined results from the coarse mask produced by Stage 1. Finally, update the segmentation head using refined results along with corresponding confidence. The proposed method demonstrates competitive performance on three common UVOS datasets (DAVIS2016, SegTrackv2, FBMS59) with higher accuracy, faster convergence, lower training cost, validating the effectiveness of our approach.

Surface defect inspection plays a vital role in the industrial production process. Many detection methods based on deep learning have been gradually applied because of their better generalization performance. However, achieving accurate annotations for training deep learning models remains a challenge due to the difficult definition of defect boundaries and the high cost of manual annotation work. Meanwhile, the detection performance of the current deep-learning methods still cannot meet the needs of industrial applications. To address these issues, this article proposes a combined anomaly aware weakly supervised lightweight model that only requires image-level labels for training and outputs defect localization. In the framework, we first design a lightweight backbone to obtain feature maps. Then, we propose a novel weakly supervised localization (WSL) method to obtain anomaly responses and use them as prior knowledge of the downstream network. Finally, the final defect detection result is obtained through the work of the designed downstream fine inspection network. In addition, we will employ multiple supervisions throughout the framework for full data use. The results of the evaluation on four real-world defect datasets demonstrate that the proposed method is superior and more generalized than state-of-the-art WSL methods and defect detection methods on average precision.