HyperTrans: Efficient Hypergraph-Driven Cross-Domain Pattern Transfer in Image Anomaly Detection

Authors: Tengyu Zhang, Deyu Zeng, Baoqiang Li, Wei Wang, Wei Liu, Zongze Wu

IJCAI 2025 | Venue PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental Extensive experiments demonstrate that Hyper Trans offers significant advantages in anomaly classification and anomaly segmentation tasks, outperforming state-of-the-art non-cross-domain methods in image-wise ROCAUC by 13%, 12%, and 15% in 1shot, 2-shot, and 5-shot settings on MVTec3D AD.
Researcher Affiliation Academia 1School of Software Engineering, Xi an Jiaotong University, 2Shenzhen University, 3Guangdong Laboratory of Artificial Intelligence and Digital Economy (SZ), 4Wuhan University EMAIL, EMAIL
Pseudocode Yes Algorithm 1 Optimization algorithm for Hyper Trans
Open Source Code Yes The domain perturbation module is integrated into layers 3 to 10 of the backbone, employing Perlin noise with a mean of 0 and a standard deviation of 0.75 for domain perturbations.1 1https://github.com/ra Rn0y/Hyper Trans
Open Datasets Yes Dataset: The MVTec-3D AD dataset [Bergmann et al., 2021] is a standard 3D anomaly detection dataset, on which our experiments are performed.
Dataset Splits Yes It consists of 10 categories, a total of 2656 training samples, and 1137 testing samples. The 3D scans were acquired by a high-resolution industrial 3D sensor using structured light. Due to the scarcity of defective samples, only normal samples are contained in the training set of MVTec-3D AD, while the testing set contains both normal and defective samples.
Hardware Specification Yes We implement Hyper Trans using Py Torch2.4.0 and conduct evaluations on an NVIDIA A100.
Software Dependencies Yes We implement Hyper Trans using Py Torch2.4.0 and conduct evaluations on an NVIDIA A100.
Experiment Setup Yes Initial values of other parameters are listed in Table 2. Table 2: Parameter settings of Hyper Trans in our experiments. ̄  ̃  ̄ epochs P epochs R 0.3 0.6 3 100