As semiconductor manufacturing advances to smaller process nodes, hotspot detection has become critical for ensuring the manufacturability and reliability of integrated circuit (IC) layouts. However, existing detection methods rely heavily on labeled data tailored to specific nodes, resulting in poor generalizability across nodes due to variations in layout geometries and fabrication processes. Labeling new data at advanced nodes is also costly and time-consuming. To overcome these challenges, we propose Node2Node, the first adaptation framework explicitly designed for cross-node hotspot detection. Node2Node integrates a novel node-invariant encoder with a node-specific encoder to jointly capture transferable and node-dependent features. To further improve robustness, we introduce a bidirectional center alignment strategy, which refines pseudo-labels by leveraging a small amount of labeled data from the target node. Additionally, a cross-node distribution loss is introduced to explicitly align feature distributions between nodes. Extensive experiments demonstrate that Node2Node substantially improves cross-node generalization and achieves state-of-the-art hotspot detection performance.

Recent advances in deep learning-based layout hotspot detection have made remarkable progress in identifying potential defect patterns at early design stages. However, most existing methods rely on supervised learning, which requires manual identification of pre-defined hotspots and leads to considerable labeling effort. Moreover, design houses often struggle to obtain a sufficient number of labeled hotspot samples, limiting the applicability and scalability of such methods. In this article, we introduce a novel approach, termed you only need non-hotspot (YONN), which to the best of our knowledge, is the first unsupervised and training-free framework for layout hotspot detection. The proposed method mitigates the dependence on labeled hotspot data by leveraging memorized prototypes and a query-based inference mechanism. Specifically, YONN employs a CNN-based prototype generation network to extract multi-scale, fine-grained representations of layouts. During inference, a combination of shape-aware and topology-aware query mechanisms facilitates precise pixel-wise matching between test layout and memorized prototypes. To further enhance YONN’s efficiency and scalability, we propose a prototype sampling strategy that integrates density-based clustering techniques, significantly reducing the scale of the prototypes. Experimental results indicate that YONN achieves performance within 10% of leading state-of-the-art supervised learning methods, despite operating in a fully unsupervised setting without access to hotspot data. As an optional extension, YONN surpasses existing state-of-the-art approaches using only 30% hotspot labels. Notably, YONN is a training-free framework that enables on-the-fly adaptation by directly incorporating novel samples into the prototype bank, thereby supporting efficient and scalable learning within design for manufacturability workflows.

Recently, machine learning-based techniques have been applied for layout hotspot detection. However, existing methods encounter challenges in capturing the decision boundary across the entire dataset and ignore the geometric properties and topology of the polygons. In this paper, we introduce CLI-HD, a novel contrastive learning framework on layout sequences and images for hotspot detection. Our framework improves the ability to distinguish between hotspots and non-hotspots by similarity computations instead of a single decision boundary. To effectively incorporate geometric information into the model training process, we propose Layout2Seq, which encodes polygon shapes as vectors within sequences that are subsequently fed into the CLIHD. Furthermore, to better represent topology information, we develop an absolute position embedding, replacing the standard position encoders used in Transformer architectures. Extensive evaluations on various benchmarks demonstrate that CLI-HD outperforms current state-of-the-art methods, with an accuracy improvement ranging from $0.82 \%$ to $4.77 \%$ and a reduction in false alarm rates by $\mathbf{4. 9 \%}$ to $\mathbf{2 3. 1 8 \%}$.

Accurate wafer defect detection is critical for yield control in advanced semiconductor manufacturing. Traditional rule-based and CNN-based methods struggle with subtle, low-contrast anomalies, while transformer detectors often suffer from high computational overhead. We propose FALCO-WAFER, a lightweight, feature-aware detector tailored for region-of-interest patches from inline AOI systems. Our architecture combines a Multi-Scale Depthwise Block for efficient texture encoding with a Token-Energy Diagonal Attention head for robust feature refinement, enabling anchor-free inference at arbitrary resolutions. Evaluated on a real-world dataset with 5,723 labeled defect images, FALCO-WAFER achieves 90.7% [email protected] and 7.19% FNR using only 13.3M parameters, outperforming both CNN and transformer baselines. Its compact design supports low-latency deployment in high-throughput inspection lines. Code is available at: https://github.com/MrJoker06/FALCO-WAFER.

Convolutional neural networks (CNNs) are frequently used to analyze remote sensing images and achieve impressive progress. Limited by the receptive field size of CNNs, small objects tended to lack adequate features to obtain more accurate segmentation results. To address this problem, we introduce a novel CNN model for coarse-to-fine segmentation called C2FNet. C2FNet comprises two stages: the coarse network and the fine network. The coarse network identifies the positions and coarse segmentation outcomes of small objects in the input image. The fine network then takes a closer look at the small objects and re-segments the patches using binary segmentation. The fine network distinguishes small objects from the background to refine small object segmentation. Finally, C2FNet employs an aggregation module that merges the binary segmentation maps and coarse outcomes to obtain accurate small object segmentation. We conducted extensive experiments on three widely accepted datasets for remote sensing image segmentation, namely the ISPRS 2-D semantic labeling Potsdam, Vaihingen, and iSAID. Our approach significantly improves the performance of baseline models, achieving a 0.24%–2.83% increase in IoU per small object class on iSAID.