馬清文

Neuron-Activity-Aware Fine-Tuning for Large Language Models: Enhancing the Sparsity–Performance Trade-off

智慧系統與應用研究所

大型語言模型、稀疏性、丟棄法、微調、神經元活躍度、Large Language Models、Sparsity、Dropout、Fine-tuning、Neuron Activity

Multi-Token Prediction with Consistency Methods: Selective Distribution Consistency + Latent Consistency Matching

智慧計算與科技研究所智慧物聯網產業碩士專班

多詞元預測、一致性學習、選擇性分布一致性、潛在一致性匹配、大型語言 模型、Multi-Token Prediction、Consistency Learning、Selective Distribution Consistency、Latent Consistency Matching、Large Language Models

Iterative Compositional Zero-Shot Generation

智慧計算與科技研究所

擴散模型、組合式條件生成、人機協作、迭代再訓練、平衡取樣、影像生成、diffusion models、Diffusion Transformer、compositional conditional generation、human-in-the-loop、iterative retraining、balanced sampling、image generation

Modulated Prior Diffusion for Medical Image Segmentation

智慧系統與應用研究所

去雜訊擴散機率模型、醫學影像分割、擴散先驗調制、條件擴散模型、輕量化模型設計、Denoising Diffusion Probabilistic Models、Medical Image Segmentation、Modulated Diffusion Prior、Conditional Diffusion Models、Lightweight Model Design

Signature Band Analysis for Sugar Content and Acidity Estimation of Fruits Using Hyperspectral Imaging with Domain Adaptation - For Examples of Solanum lycopersicum

智慧系統與應用研究所・共同指導：歐陽盟

高光譜影像、平均積分梯度、特徵波段、番茄、糖酸比、糖度估測、酸度估測、Hyperspectral imaging、mean integrated gradients、signature band、Solanum lycopersicum、sugar-acid ratio、soluble solids content estimation、titratable acidity estimation

Application of Synthetic Defect Image Generation to Improve PCB Inspection Accuracy and Classification Performance in SMT Manufacturing

智慧與綠能產學研究所・共同指導：歐陽盟

生成式模型、生成瑕疵影像、深度學習、印刷電路板、影像分類、機器學習、Generative model、Synthesize Defect Images、Printed Circuit Board、Surface Mount Technology、Machine Learning

Compositional Conditional Consistency Models

智慧系統與應用研究所

圖像生成、擴散模型、一致性模型、組合式零樣本圖像生成、擴散模型蒸餾訓練、Image Generation、Diffusion Models、Consistency Models、Compositional Zero- shot Generation、Diffusion Model Distillation

Handling Indeterminate Predictions of Vascular Access Dysfunction in Chronic Hemodialysis Patients through Tree-Based Approaches

智慧系統與應用研究所

血管通路功能不全、樹模型、機器學習、不確定性量化、Vascular Access Dysfunction、Tree-Based Models、Machine Learning、Indeterminacy Quantification

Defect Classification with Indeterminate Awareness Using Gaussian Process and Cross-Entropy Hybrid Models

智慧與綠能產學研究所

瑕疵分類、混合專家、不定性分析、深度學習、高斯過程、Defect Classification、Hybrid Expert、Indeterminate Analysis、Deep Learning、Gaussian Process

Enhancing Deep Learning Model Performance with Memory Bank: A Novel Approach to Hard Example Mining

智慧與綠能產學研究所・共同指導：曾煜棋

記憶銀行、目標檢測、分類任務、增量式學習、主動式學習、Memory Bank、Object Detection、Classification Task、Incremental Learning、Active Learning

Unseen defect image synthesis with compositional conditional diffusion model

智慧計算與科技研究所

深度學習、擴散模型、瑕疵檢測、圖像生成、組合零樣本學習、Deep Learning、Diffusion Model、Defect Detection、Image Generation、Compositional Zero-Shot Learning (CZSL)

Compositional Class label-to-image Diffusion Model

智慧系統與應用研究所

圖像生成、擴散模型、組合式零樣本圖像生成、組合式零樣本學習、Image generation、Diffusion Model、Compositional zero-shot image generation、Compositional zero-shot learning iv

Partial Compositional Zero-Shot Learning by Uncertainty-Accuracy Mapping and Hierarchical Mixture of Experts

智慧與綠能產學研究所

集成學習、不確定性、組合零次學習、專家混合、模型較準、Ensemble learning、Uncertainty、Compositional Zero-Shot Learning、Mixture of Experts、Model Calibration

New Component Defect Detection Using Deep Learning and Gaussian Processes in Industrial Applications with Uncertainty Estimation

智慧系統與應用研究所

深度學習、高斯過程、異常檢測、瑕疵檢測、不確定性預測、Deep learning、Gaussian process、Anomaly detection、Defect detection、Uncertainty estimation

Unknown-aware Defect Detection of New Components in Printed Circuit Board

智慧與綠能產學研究所

瑕疵檢測、度量學習、未知類別偵測、深度學習、高斯混合模型、Defect detection、Metric learning、Unknown detection、Deep learning、Gaussian Mixture Model

Stereo Object Detection with Sparse LiDAR

智慧與綠能產學研究所

3D 物件辨識、深度估計、虛擬光學雷達、多感測器特徵融合、特徵強化、3D object detection、Depth estimation、Pseudo LiDAR、Multi-sensor feature fusion、feature enhancement

Growth Area Estimation of Bolboschoenus planiculmis in Gaomei Wetland Using Pyramid Scene Parsing Network on UAV Images

智慧與綠能產學研究所

金字塔場景分析網路、深度學習、神經網路、影像辨識、語義分割、PSPNet、Deep Learning、Neural Network、Image Segmentation、Semantic Segmentation

Multi-Scale Attention Learning for Semiconductor Wafer Defect Classification

智慧與綠能產學研究所

晶圓瑕疵檢測、瑕疵分類、多尺度注意力分析、細粒度分析、長尾資料集、Wafer defect detection、Defect classification、Multi-scale attention learning、Fine-grained visual classification、Long-tail dataset