Chengbin Peng

In surface quality inspection of optical module housings, deep learning-based automated optical inspection systems often suffer from insufficient training data due to the rarity and long-tailed distribution of defect samples, which introduces significant challenges and limits the reliability of the inspection system. To address this challenge, we propose Inpainting Diffusion-based Data Augmentation (IDDA), a novel data augmentation framework that integrates defect embedding learning and diffusion-based inpainting to enhance the performance and robustness of visual defect detection systems. First, we freeze the UNet and optimize the text encoder to extract defect-specific semantic representations from box-level annotations, creating a calibrated defect descriptor. Then, we introduce a mask-Guided denoising U-Net adaptation, fine-tuning the model to synthesize visually realistic and semantically coherent defects within specified regions of an image, ensuring both spatial and semantic coherence with the context. Experimental results on our dedicated Optical Module Housing Surface Defect (OMHSD) benchmark demonstrate that the defect samples synthesized by our method exhibit superior texture fidelity and global coherence compared to traditional generative approaches. More importantly, by integrating IDDA-generated data, the detection system shows markedly improved detection accuracy and generalization capability, effectively reducing the reliance on manual annotation and offering a pragmatic solution to data scarcity in precision surface inspection.

In fine-grained road scene understanding, semantic segmentation plays a crucial role in enabling vehicles to perceive and comprehend their surroundings. By assigning a specific class label to each pixel in an image, it allows for precise identification and localization of detailed road features, which is vital for high-quality scene understanding and downstream perception tasks. A key challenge in this domain lies in improving the recognition performance of minority classes while mitigating the dominance of majority classes, which is essential for achieving balanced and robust overall performance. However, traditional semi-supervised learning methods often train models overlooking the imbalance between classes. To address this issue, firstly, we propose a general training module that learns from all the pseudo-labels without a conventional filtering strategy. Secondly, we propose a professional training module to learn specifically from reliable minority-class pseudo-labels identified by a novel mismatch score metric. The two modules are crossly supervised by each other so that it reduces model coupling which is essential for semi-supervised learning. During contrastive learning, to avoid the dominance of the majority classes in the feature space, we propose a strategy to assign evenly distributed anchors for different classes in the feature space. Experimental results on multiple public benchmarks show that our method surpasses traditional approaches in recognizing tail classes.

Semi-Supervised Domain Adaptation (SSDA) leverages knowledge from a fully labeled source domain to classify data in a partially labeled target domain. Due to the limited number of labeled samples in the target domain, there can be intrinsic similarity of classes in the feature space, which may result in biased predictions, even when the model is trained on a balanced dataset. To overcome this limitation, we introduce a multi-view consistency framework, which includes two views for training strongly augmented data. One is a debiasing strategy for correcting class-wise prediction probabilities according to the prediction performance of the model. The other involves leveraging pseudo-negative labels derived from the model predictions. Furthermore, we introduce a cross-domain affinity learning aimed at aligning features of the same class across different domains, thereby enhancing overall performance. Experimental results demonstrate that our method outperforms the competing methods on two standard domain adaptation datasets, DomainNet and Office-Home. Combining unsupervised domain adaptation and semi-supervised learning offers indispensable contributions to the industrial sector by enhancing model adaptability, reducing annotation costs, and improving performance.

Link prediction is a classical problem in graph analysis with many practical applications. For directed graphs, recently developed deep learning approaches typically analyze node similarities through contrastive learning and aggregate neighborhood information through graph convolutions. In this work, we propose a novel graph neural network (GNN) framework to fuse feature embedding with community information. We theoretically demonstrate that such hybrid features can improve the performance of directed link prediction. To utilize such features efficiently, we also propose an approach to transform input graphs into directed line graphs so that nodes in the transformed graph can aggregate more information during graph convolutions. Experiments on benchmark datasets show that our approach outperforms the state-of-the-art in most cases when 30%, 40%, 50%, and 60% of the connected links are used as training data, respectively.