Dr. Guoyang Xie(谢国洋)

I am an Algorithm Manager in the Department of Intelligent Manufacturing, CATL . I achieved Machine Learning PhD degree at University of Surrey, NICE Group , supervised by Prof. Yaochu Jin. My research focus on AI for Manufacturing and Batteries.

Email  /  CV  /  中文版简历  /  Google Scholar  /  Github  /  微信号：xgy_cn

Prospective Postdocs & Enginneers & Interns & PhDs: I am always looking for self-motivated persons with strong mathematical and programming background. If you are interested with working with me, please feel free to drop me an email.

Now I'm interested in AI for manufacturing and batteries. Previously, much of my research is about detecting and localizing anomalies for both industrial images and medical images. Representative papers are highlighted.

Note that *contributed equally, #corresponding author

We introduce the Multi-View Projection (MVP) framework, leveraging pre-trained Vision-Language Models (VLMs) to detect anomalies.

We propose a novel group-level feature-based network, called Group3AD, to address 3D anomaly detection task.

We propose a mutltiple thresholding (Multi-T) module to improve outlier scoring methods.

We introduce multi-modality anomaly synthesis model to generate more logical anomalies and propose a new multi-modality AD dataset (MVTec-Caption) to examine the performance of multi-modality AD.

We present a simple and effective mask-augmented patch embedding to integrate shadow information and promote the model’s emphasis on acquiring knowledge for shadow regions.

My PhD Final Thesis. There are six chapters, including Introduction, IAD Taxonomy, IM-IAD, GraphCore, TransferAD and Future Work

We introduce a 3D point cloud dataset for industrial anomaly detection.

We reveal that rotation-invariant feature property has a significant impact in industrial-based fewshot anomaly detection.

We propose a large-scale systematic benchmark and uniform setting for IAD to bridge the gap between academy and industrial manufacturing

We provide a comprehensive review of deep learning-based IAD from the perspectives of neural network architectures, levels of supervision, loss functions, metrics and datasets.

We propose a multi-modality reconstruction-based network for RGBD AD, which eliminate the usage of memory bank and pretrained model. Moreover, the proposed method obtains the best trade-off between the accuracy and inference speed.

We systematically investigate various data augmentation methods for few-shot IAD algorithms.

We proposed a method that reducing the demands for deformable registration while encouraging to leverage the misaligned and unpaired data

We provide a comprehensive review of cross-modality synthesis for neuroimages, from the perspectives of weakly-supervised and unsupervised settings, loss functions, evaluation metrics, ranges of modality, datasets, and the synthesis-based downstream applications.

We proposed a new benchmark for federated domain translation on unsupervised brain image synthesis to bridge the gap between federated learning and medical GAN.

We propose a multi-objective oneshot network architecture search algorithm to obtain the best trade-off networks in terms of the adversarial accuracy, the clean accuracy and the model size.

We propose a novel metric, K-CROSS, to evaluate the quality of cross-modality synthesized neuroimage

We proposed a new algorithm for online extrinsic calibration of multi-LRFs by observing a planar checkerboard pattern and solving for transformation between the views of a planar checkerborard from a camera and multi-LRF.

Feel free to steal this website's source code. Thanks to JonBarron.