政府計畫（GRB），建議「依年度遞減排序」，以查看最新的研究方向。
| 畢業學年度 | 論文標題 | 連結 | 學位 | 畢業時長(年) |
|---|---|---|---|---|
| 關鍵字 | 口試日期 | |||
| 114 | 輕量化稀疏... 輕量化稀疏感知手部姿態模型 (SPLite Hand: Sparsity-Aware Lightweight 3D Hand Pose Estimation) | NTHU NDLTD | 碩 | 2.31 |
模型輕量化設計 (light-weight model design)、稀疏卷積 (sparse convolution)、手部姿態估計 (hand pose estimation)、量化感知訓練(quantization-aware training) 模型輕量化... | 2025-11-19 | |||
| 114 | 三維場景中... 三維場景中的部件層級實例分割 (PS3: Part level instance segmentation in 3D) | NTHU NDLTD | 碩 | 2.29 |
開放詞彙 (Open Vocabulary)、三維實例分割 (3D Instance Segmentation)、部件層級 (Part-level)、階層式特徵表示 (Hierarchical Feature Representation)、免訓練(Training-Free) 開放詞彙 ... | 2025-11-12 | |||
| 114 | 具定位感知... 具定位感知能力的令牌剪枝：對抗令牌裁減導致的視覺定位退化 (Grounding-Aware Token Pruning: Addressing Visual Grounding Degradation from Token Reduction) | NTHU NDLTD | 碩 | 1.98 |
. .... | 2025-07-23 | |||
| 114 | 利用稀疏表... 利用稀疏表示提升卷積與轉換器之效率 (Leveraging Sparse Representation for Convolution and Transformer) | NTHU NDLTD | 碩 | 2.02 |
稀疏卷積(Sparse Convolution)、人體姿態估計(Human Pose Estimation)、稀疏表示(Sparse Representation)、激活函數稀疏性(Activation Sparsity) 稀疏卷積(... | 2024-08-06 | |||
| 113 | 將激活稀疏... 將激活稀疏性與詞元層級稀疏性結合於 ViT 模型 (Activation Sparsity Meets Token-Level Sparsity in ViT) | NTHU NDLTD | 碩 | 1.99 |
. .... | 2025-07-28 | |||
| 113 | 模組化物件... 模組化物件導向的三維問答方法 (POC-QA: Plug-and-Play Object-Centric 3D Question Answering) | NTHU NDLTD | 碩 | 1.94 |
. .... | 2025-07-09 | |||
| 113 | P2-OV... P2-OV3D：具備模組化插拔設計的開放詞彙三維感知系統 (P2-OV3D: Plug-and-Play Open-Vocabulary Perception in 3D) | NTHU NDLTD | 碩 | 1.94 |
. .... | 2025-07-07 | |||
| 113 | 開放詞彙之... 開放詞彙之無標註多視角室內3D物件辨識 (OpenM3D: Open Vocabulary Multi-view Indoor 3D Object Detection without Human Annotations) | NTHU NDLTD | 碩 | 2.4 |
開放詞彙(Open-Vocabulary)、3D物件辨識(3D Object Detection)、多視角(Multi-View Image) 開放詞彙(... | 2024-12-25 | |||
| 113 | 基於大語言... 基於大語言模型預訓練及資料增強演算法提升電子鼻在跨域場域中少樣本的肺癌辨識能力 (Leveraging Large Pre-trained Language Models and Data Augmentation Algorithms to Enhance eNose Lung Cancer Recognition in Cross-Domain Few-Shot Scenarios) | NTHU NDLTD | 碩 | 2.25 |
肺癌(Lung cancer)、呼吸檢測(breath tests)、電子鼻(eNose)、多種資料擴增演算法(various data augmentation algorithms)、LLM(LLM)、遷移學習(transfer learning)、跨域應用(cross-domain) 肺癌(Lu... | 2024-10-28 | |||
| 113 | 多視角一致... 多視角一致性用於自監督學習的房間佈局幾何 (Multi-view Consistency for Self-training Room Layout Geometries) | NTHU NDLTD | 博 | 7.22 |
全景影像(360 images)、半監督式學習(semi-supervised learning)、多視角幾何(multi-view geometry) 全景影像(... | 2024-10-18 | |||
| 113 | 稀疏模態在... 稀疏模態在人體與手部姿態辨識中的隱私與效能 (Using Sparse Modalities in Human and Hand Pose Estimation for Privacy and Efficiency) | NTHU NDLTD | 碩 | 2.12 |
. .... | 2024-09-11 | |||
| 113 | 室內場景理... 室內場景理解的全景結構交換 (PanoMixSwap – Panorama Mixing via Structural Swapping for Indoor Scene Understanding) | NTHU NDLTD | 碩 | 2.07 |
. .... | 2024-08-23 | |||
| 112 | 工業物件偵... 工業物件偵測應用的高效模型部署與訓練策略 (Efficient Model Deployment and Training Strategies for Industrial Object Detection Applications) | NTHU NDLTD | 碩 | 1.99 |
物件檢測(ObjectDetection)、光流估計(OpticalFlowEstimation)、資料增強(DataAugmentation)、主動採樣(ActiveSample)、量化(Quantization)、提示工程(PromptEngineering) 物件檢測(... | 2024-07-26 | |||
| 112 | 使用擴散模... 使用擴散模型的有效室內全景資料擴增方法研究 (Controlled Panorama Data Augmentation using Diffusion Models) | NTHU NDLTD | 碩(外籍生) | 2.85 |
全景圖像(360)、深度學習(deep-learning)、語義分割(semantic-segmentation)、佈局估計(layout-estimation)、數據增強(diffusion-models) 全景圖像(... | 2024-06-05 | |||
| 112 | 利用擴散模... 利用擴散模型的稀疏視圖無相機姿態重建 (Inverting Diffusion for Pose-Free Reconstruction from Sparse Views) | NTHU NDLTD | 碩 | 2.61 |
三維重建(3D Reconstruction)、三維生成(3D Generation)、相機位姿估計(Camera Pose Estimation) 三維重建(... | 2024-03-11 | |||
| 111 | 從 360... 從 360 全景影像重建不同細節層次之室內幾何 (Indoor Geometry Reconstruction from 360 to Different Levels) | NTHU NDLTD | 碩逕博 | 4.99 |
360(360)、全景影像(panorama)、室內幾何重建(indoor geometry reconstruction)、房間格局(room layout)、平面偵測(plane detection)、深度預測(depth estimation)、重力對齊(gravity alignment) 360(3... | 2023-07-27 | |||
| 111 | 結合高速公... 結合高速公路於終身多智能體路徑規劃之研究 (The Study of Highway for Lifelong Multi-Agent Path Finding) | NTHU NDLTD | 碩 | 1.94 |
路徑規劃(Path Planning)、多智能體路徑規劃(Multi-Agent Path Finding)、終身多智能體路徑規劃(Lifelong Multi-Agent Path Finding)、人工智慧(Artificial Intelligence)、多機器人系統(Multi-Robot Systems) 路徑規劃(... | 2023-07-10 | |||
| 111 | AI 輔助... AI 輔助效率分析之影像辨識系統 (AI-assisted Efficiency Analysis for Image Recognition System) | NTHU NDLTD | 碩 | 1.94 |
AI輔助(AI-assisted)、效率分析(Efficiency Analysis)、影像辨識系統(Image Recognition System) AI輔助(... | 2023-07-10 | |||
| 111 | 基於跨模態... 基於跨模態提升匹配影片和音樂 (Video and Music Matching via Cross-Modality Lifting) | NTHU NDLTD | 碩 | 1.94 |
跨模態(Cross-Modality)、度量學習(Metric Learning)、匹配系統(Matching System)、短影片配樂(Short Video Background Music)、深度學習(Deep Learning) 跨模態(C... | 2023-07-10 | |||
| 111 | 運用混合式... 運用混合式量子與經典深度學習於半導體晶圓缺陷自動識別 (Automatic Identification of Semiconductor Defect Using Hybrid Quantum-Classical Deep Learning) | NTHU NDLTD | 博 | 3.70 |
深度學習(Deep Learning)、量子神經網路(Quantum Neural Network)、缺陷檢測(Defect Detection) 深度學習(... | 2023-04-12 | |||
| 111 | 全景室內深... 全景室內深度與格局偵測 (360 Perception for Indoor Depth and Layout Estimation) | NTHU NDLTD | 博 | 3.57 |
全景影像(360)、深度學習(deep-learning)、深度預測(depth-estimation)、格局預測(layout-estimation) 全景影像(... | 2023-02-22 | |||
| 111 | 建造短影音... 建造短影音背景音樂推薦系統的困難 (Difficulties in building a short-video background music recommendation system) | NTHU NDLTD | 碩(提早入學) | 3.00 |
推薦系統(Recommendation System)、跨模態匹配(Cross Modal Matching)、音樂推薦(Music Recommendation)、短影音背景音樂推薦(Short Video Background Music Recommendation)、深度學習(Deep Learning)、表徵學習(Representation Learning) 推薦系統(... | 2023-01-31 | |||
| 111 | 醫療影像與... 醫療影像與其他資訊之結合應用 (Medical Image + X Application) | NTHU NDLTD | 碩(提早入學) | 1.95 |
生成對抗模型(GAN)、圖像對圖像轉換(Image-to-Image Translation)、電腦斷層圖(CT)、圖像上色(Image Colorization)、跨模態(Cross-modality)、半監督式學習(Semi-supervised Learning)、胸腔X光圖(Chest X-ray Image)、電子鼻(Electronic Nose) 生成對抗模... | 2023-01-11 | |||
| 111 | 基於光流預... 基於光流預測及物件偵測模型的主動式取樣方法 (Active Sampling based on the optical flow prediction combined with object detection model) | NTHU NDLTD | 碩 | 2.09 |
主動式取樣(Active Sampling)、光流預測(Optical flow prediction)、物件偵測(Object detection) 主動式取樣... | 2022-09-01 | |||
| 111 | 利用人物移... 利用人物移動軌跡之分群降低產生日常活動狀況的標註成本 (Reduce the Annotation Effort to Generate Activities of Daily Living by Human Tracklet Clustering) | NTHU NDLTD | 碩 | 2.09 |
電腦視覺(Computer Vision)、機器學習(Machine Learning)、移動軌跡分群(Human Tracklet Clustering) 電腦視覺(... | 2022-09-01 | |||
| 110 | 使用基於注... 使用基於注意力模型的特徵聚合用於每個像素的密集預測 (Dense Prediction with Attentive Feature Aggregation) | NTHU NDLTD | 碩 | 1.87 |
密集預測(Dense Prediction)、特徵聚合(Feature Aggregation)、深度學習(Deep Learning)、電腦視覺(Computer Vision) 密集預測(... | 2022-06-14 | |||
| 110 | 基於 2D... 基於 2D 模型的知識轉換減少對 3D 標註需求之訓練策略 (Data Efficient 3D Learner via Knowledge Transferred from 2D Model) | NTHU NDLTD | 碩 | 1.87 |
語意分割(segmentation)、小數據學習(low-shot learning)、場景分析(scene analysis and understanding)、深度學習(deep learning) 語意分割(... | 2022-06-14 | |||
| 110 | 實時偵測器... 實時偵測器之在不同場域的效果差距 (UnGap Real-time Detectors by Orchestrating the Self-Training of Teacher and Students) | NTHU NDLTD | 碩 | 1.87 |
物件偵測(Object detection)、無監督領域自適應(Unsupervised domain adaptation)、半監督學習(Semi-supervised learning) 物件偵測(... | 2022-06-14 | |||
| 110 | 類別層級之... 類別層級之關節式形變類神經輻射場 (CLA-NeRF: Category-Level Articulated Neural Radiance Field) | NTHU NDLTD | 碩 | 2.58 |
深度學習(Deep Learning)、新穎視角合成(Novel-View Synthesis)、類神經輻射場(NeRF)、關節式物件(Articulated Object)、電腦視覺(Computer Vision) 深度學習(... | 2022-02-25 | |||
| 110 | 跨越人類感... 跨越人類感知:代理人的 360 視角導航與深度感知 (Beyond Human Vision: 360 View Pilot and Depth Perception for an Embodied Agent) | NTHU NDLTD | 碩逕博 | 6.25 |
單目三維物體追蹤(Monocular 3D Object Tracking)、準稠密個體相似度(Quasi Dense Instance Similarity)、長短期記憶模型(Long Short Term Memory)、360 度影片(360 degree Videos)、注意力輔助系統(Focus Assistance System) 單目三維物... | 2021-10-29 | |||
| 109 | 人體動作辨... 人體動作辨識系統的弱點分析與對策 (Troubleshooting in Human Action Recognition) | NTHU NDLTD | 碩 | 2.19 |
機器學習(Machine learning)、人工智慧(Data augmentation)、資料擴增(Feature engineering)、特徵工程(undefined) 機器學習(... | 2021-10-08 | |||
| 109 | 透過實例感... 透過實例感知參數化減少在線持續學習時的災難性遺忘 (Mitigating Forgetting in Online Continual Learning via Instance-Aware Parameterization) | NTHU NDLTD | 碩 | 1.90 |
深度學習(Deep Learning)、神經網路架構探索(Neural Architecture Search)、持續學習(Continual Learning)、在線學習(Online Learning) 深度學習(... | 2021-06-25 | |||
| 109 | 從電腦斷層... 從電腦斷層圖到全彩圖片：使用二階段生成對抗模型實現電腦斷層圖著色 (CT2RGB: Using the Two-stage Generative Adversarial Network to Colorize Computed Tomography Images) | NTHU NDLTD | 碩 | 1.90 |
生成對抗模型(GAN)、圖像對圖像轉換(Image-to-Image Translation)、電腦斷層圖(CT)、圖像上色(Image Colorization) 生成對抗模... | 2021-06-25 | |||
| 109 | 實例感知神... 實例感知神經架構搜索 (InstaNAS: Instance-aware Neural Architecture Search) | NTHU NDLTD | 碩 | 2.59 |
神經架構搜索(Neural Architecture Search) 神經架構搜... | 2021-03-02 | |||
| 108 | 基於魚眼鏡... 基於魚眼鏡頭下的人類行為分析的日常摘要生成系統 (A Daily Summary Generating System via Fisheye-based Human Behavior Analysis) | NTHU NDLTD | 碩 | 1.96 |
電腦視覺(Computer Vision)、魚眼影像(Fisheye Image)、行為分析(Behavior Analysis) 電腦視覺(... | 2020-07-17 | |||
| 108 | 基於隱私保... 基於隱私保護和資料有效性的魚眼居家行為分析系統 (A Privacy Preserved Behavior Analysis System with Data Efficient Strategy via Fisheye Camera) | NTHU NDLTD | 碩 | 1.96 |
行為分析(Behavior Analysis)、居家照護(Home Care)、隱私保護(Privacy Preserving)、電腦視覺(Computer Vision)、資料有效性(Data Efficient) 行為分析(... | 2020-07-17 | |||
| 108 | 應用變分自... 應用變分自編碼器強化物件偵測的資料篩選效能 (Enhance data selection efficiency with variational auto-encoder for object detection’s active learning) | NTHU NDLTD | 碩(提早入學) | 2.44 |
物件偵測(Object Detection)、自主學習(Active Learning)、電腦視覺(Computer Vision)、變分解碼器(Variational Auto-encoder) 物件偵測(... | 2020-07-10 | |||
| 108 | 360SD... 360SD-Net: 360度雙目深度估測與可學習立體成本容積 (360SD-Net: 360° Stereo Depth Estimation with Learnable Cost Volume) | NTHU NDLTD | 碩(提早入學) | 2.44 |
360度影像(360° Image)、雙目深度估測(Stereo Depth Estimation)、全景影像(Equirectangular Image)、卷積神經網路(Neural Network)、深度學習(Deep Learning)、電腦視覺(Computer Vision) 360度影... | 2020-07-10 | |||
| 108 | 用光達資料... 用光達資料增進單目和雙目深度預測 (Improve monocular and stereo depth estimation with LiDAR data) | NTHU NDLTD | 碩 | 2.64 |
深度預測(depth estimation)、單目(monocular)、雙目(stereo)、光達(LiDAR) 深度預測(... | 2020-03-20 | |||
| 108 | 基於深度學... 基於深度學習與震動反饋的可穿戴電腦視覺系統於視障者輔助之應用 (Deep Learning-based Wearable Vision-system with Vibrotactile-feedback for Visually Impaired People to Reach Objects) | NTHU NDLTD | 碩 | 2.64 |
視障輔助(Blind and Visually Impaired assistace)、物體偵測(Object detection)、影像辨識(Image recognition)、深度學習(Deep learning)、互動設計(Interaction design)、即時系統(Real-time system) 視障輔助(... | 2020-03-20 | |||
| 108 | 卷積閘控圖... 卷積閘控圖神經網路應用於放射治療計畫輪廓勾畫 (Radiotherapy Target Contouring with Convolutional Gated Graph Neural Network) | NTHU NDLTD | 碩 | 2.45 |
放射治療計畫輪廓(Radiotherapy Target Contour)、影像分割(Image Segmentation)、圖神經網路(Graph Neural Network) 放射治療計... | 2020-01-10 | |||
| 108 | 基於手指穿... 基於手指穿戴式相機和聽覺回饋於室內空間之視障者尋物導引系統 (RingGuardian: A finger-worn Camera-based System for Blind and Visually Impaired Users to Perform Room-level Search of Objects with Audio Guidance) | NTHU NDLTD | 碩 | 3.18 |
視障輔助(Blind and Visually Impaired assistace)、物體偵測(Object detection)、影像辨識(Image recognition)、深度學習(Deep learning)、互動設計(Interaction design)、即時系統(Real-time system) 視障輔助(... | 2019-10-04 | |||
| 107 | 基於自我監... 基於自我監督式學習360影片之深度與相機位移 (Self-Supervised Learning of Depth and Camera Motion from 360 Videos) | NTHU NDLTD | 碩 | 1.93 |
全景(panorama)、深度(depth)、深度學習(deep-learning)、人工智慧(ai)、計算機視覺(cv)、點雲(point-cloud) 全景(pa... | 2019-07-05 | |||
| 107 | 利用動作資... 利用動作資訊改善跨域人體目標分割 (Leveraging Motion Information for Improving Domain Adaptation in Human Segmentation) | NTHU NDLTD | 碩(外籍生) | 2.47 |
域適應學習(Domain Adaptation)、人體目標分割(Human Segmentation)、動作資訊(Motion Information) 域適應學習... | 2019-01-18 | |||
| 107 | 基於電腦斷... 基於電腦斷層影像之食道癌放射治療計劃劑量區域擷取系統 (Using Computed Tomography Images for Dose Contour Segmentation in Esophageal Cancer Radiotherapy Treatment Plan) | NTHU NDLTD | 碩 | 2.39 |
食道癌(Esophageal Cancer)、電腦斷層影像(Computed Tomography)、劑量區域切割(Dose Contour Segmentation) 食道癌(E... | 2018-12-21 | |||
| 106 | 基於影片中... 基於影片中移動資訊改善跨域人體目標分割 (Leveraging Motion Priors in Videos for Improving Human Segmentation across Domains) | NTHU NDLTD | 碩 | 1.91 |
主動學習(Active learning)、域適應學習(Domain Adaptation)、人體目標分割(Human segmentation) 主動學習(... | 2018-06-28 | |||
| 106 | 基於多模態... 基於多模態系統預測人類意圖與提示遺忘事件之持續學習 (A Multimodal System Towards Intention Anticipation and Missing Actions Reminder via Continuous Self-Learning) | NTHU NDLTD | 碩 | 1.91 |
意圖預測(Intention Anticipation)、遺忘事件提醒(Missing Action Reminder)、智慧家庭(Smart Home)、融合感測器(Sensor Fusion)、連續學習(Continuous Learning) 意圖預測(... | 2018-06-28 | |||
| 106 | 以不一致性... 以不一致性損失函數結合抽取式和生成式摘要的融合摘要模型 (A Unified Model for Extractive and Abstractive Summarization using Inconsistency Loss) | NTHU NDLTD | 碩 | 1.91 |
自然語言處理(NLP)、摘要(summarization)、深度學習(deeplearning) 自然語言處... | 2018-06-27 | |||
| 106 | 高效率不確... 高效率不確定性預測應用於影像語意分割 (Efficient Uncertainty Estimation for Semantic Segmentation in Videos) | NTHU NDLTD | 碩 | 1.91 |
不確定性(uncertainty)、語意分割(segmentation)、影像(video)、高效率(efficient) 不確定性(... | 2018-06-27 | |||
| 106 | 立方填補於... 立方填補於360影片之非監督式學習 (Cube Padding for Unsupervised Saliency Prediction in 360 Videos) | NTHU NDLTD | 碩 | 2.85 |
360全景影片(360 Videos)、立方填補(Cube Padding)、非監督式學習(Unsupervised Learning)、顯著度預測(Saliency Prediction)、深度類神經網路(Deep Neural Network) 360全景... | 2018-06-04 | |||
| 106 | 室內機器人... 室內機器人輔助導向之基於深度學習之俯瞰視角物品偵測 (Deep Learning based Bird's Eye View Detection toward Indoor Robot Assistant System) | NTHU NDLTD | 碩 | 2.51 |
目標偵測(Object detection)、深度學習(Deep learning)、鳥瞰偵測(Bird's eye view deteciton)、室內物品偵測(Indoor objects detection) 目標偵測(... | 2018-02-02 | |||
| 106 | 基於多模態... 基於多模態資料訓練人類意圖預測及液體傾倒監測 (Anticipating Human Intention and Monitoring Liquid Pouring via Multimodal Learning) | NTHU NDLTD | 碩(提早入學) | 1.98 |
深度學習(deep learning)、多模態資料(multimodal data)、增強式學習(reinforcement learning) 深度學習(... | 2018-01-23 | |||
| 106 | 基於行車紀... 基於行車紀錄器提取駕駛行為-行車全球座標定位 (Extracting Driving Behavior: Global Metric Localization from Dashcam Videos in the Wild) | NTHU NDLTD | 碩 | 2.37 |
駕駛行為(driving behavior)、全球定位(global localization)、行車紀錄器(dashcam) 駕駛行為(... | 2017-12-13 | |||
| 105 | 影片標題產... 影片標題產生與問答 (Video titling and Question-Answering) | NTHU NDLTD | 碩 | 2.93 |
電腦視覺(CV)、深度學習(DL)、遞迴式神經網路(RNN)、影片(Video)、標題(Title)、問答(Question-Answering) 電腦視覺(... | 2017-07-06 | |||
| 105 | 基於深度學... 基於深度學習預測交通意外及事故 (Anticipating Accidents based on Deep Learning in Dashcam Videos) | NTHU NDLTD | 碩 | 1.93 |
電腦視覺(Computer Vision)、機器學習(Machine Learning)、深度學習(Deep Learning) 電腦視覺(... | 2017-07-06 | |||
| 105 | 基於深度學... 基於深度學習從手部視角辨識影像 (Recognition from Hand Cameras: A Revisit with Deep Learning) | NTHU NDLTD | 碩 | 1.93 |
深度學習(deep learning)、手部視角(hand camera)、影像辨識(image recognition) 深度學習(... | 2017-07-06 | |||
| 105 | 基於對抗式... 基於對抗式訓練生成跨域影像描述 (Generating Cross-domain Visual Description via Adversarial Learning) | NTHU NDLTD | 碩 | 1.93 |
深度學習(Deep Learning)、圖像字幕生成(Image Captioning)、遷移學習(Transfer Learning)、電腦視覺(Computer Vision)、對抗式訓練(Adversarial Training)、增強學習(Reinforcement Learning) 深度學習(... | 2017-07-06 | |||
| 105 | 基於對抗式... 基於對抗式訓練的跨城市街景分割 (Cross City Adaptation of Road Scene Segmenters via Adversarial Learning) | NTHU NDLTD | 碩 | 1.93 |
街景語意分割(Road-Scene-Semantic-Segmentation)、域適應性(Domain-Adaptation)、對抗式訓練(Adversarial-Learning) 街景語意分... | 2017-07-06 | |||
| 104 | 影像視覺的... 影像視覺的定位研究 (The Study of Vision Based Positioning) | NTHU NDLTD | 碩 | 無口試日期 |
影像定位(vision based posotioning)、室內定位(indoor positioning) 影像定位(... | ||||