鄭文皇

Semantic Knowledge Learning for Remote Sensing Image Understanding: Learning, Representation, and Accumulation of Semantic Knowledge

資訊科學與工程研究所・共同指導：黃敬群

遙測影像理解、基礎模型、弱監督語意分割、持續學習、Remote Sensing Image Understanding、Foundation Vision–Language Models、Semantic Knowledge Learning、Weakly Supervised Semantic Segmentation、Continual Learning

Leveraging Knowledge Distillation on Multi-Exit Vision Transformer for Small-Data Image Classification

電機資訊國際博士學位學程・共同指導：黃敬群

視覺轉換器、自組學習、殘留標記嵌入、影像分類、知識提煉、Vision transformer、Self-ensemble learning、Residual token embedding、Image classification、Knowledge distillation

Multimodal Large Language Model Guided Diffusion for Visual Synthesis and Editing

電機學院博士班・共同指導：帥宏翰

擴散模型、多模態大語言模型、多模態學習、影像生成、影像編輯、Diffusion Models、Multimodal Large Language Models、Multimodal Learning、Image Generation、Image Editing

Visual Intelligence for Human-Centered AI

電子研究所・共同指導：簡鳳村

人本導向視覺智慧、臉部表情辨識、微表情辨識、穩健視覺感知、影像免疫、可控影片生成、Human-Centered Visual Intelligence、Facial Expression Recognition、Micro Expression Recognition、Robust Visual Perception、Image Immunization、Controllable Video Generation

Evolving Frameworks for Object Detection: From Tiny Targets to 3D Scenes

電機工程學系・共同指導：帥宏翰

小物件辨識、單目三維物件辨識、Tiny Object Detection、Monocular 3D Object Detection

Coarse-to-fine Refinement for Fine-grained Visual Recognition

電機工程學系・共同指導：帥宏翰

細粒度視覺分類、視覺情感識別、多模態大語言模型、細粒度分佈外檢測、Fine-grained Visual Classification、Visual Emotion Recognition、Multi-modal Large Language Models、Fine-grained Out-of-distribution Detection

Enhancing Contrastive Learning Frameworks for Sequential Recommendation Systems with Hard Negatives and Soft Labels

電子研究所・共同指導：簡鳳村

序列推薦、推薦系統、對比學習、困難負樣本、軟標籤、sequential recommendation、recommendation system、contrastive learning、hard negatives、soft label

A Taiwanese Street Scene Data Augmentation Method Based on Diffusion Model for Layout-to-Image Generation

電子研究所・共同指導：簡鳳村

擴散模型、物件偵測、台灣街景、上下文學習、diffusion model、object detection、Taiwan street scene、in-context learning

Enhancing Coherence and Fidelity in Camouflaged Image Generation via Foreground Focus

人工智慧技術與應用碩士學位學程・共同指導：帥宏翰

偽裝圖像生成、潛在擴散模型、前景聚焦、和諧性增強、保真度維持、Camouflaged Image Generation、Latent Diffusion Model、Foreground Focus、Coherence Enhancement、Fidelity Preservation

A Universal Framework for Heterogeneous Knowledge Distillation based on Region-Aware Attention and Adaptive Feedback Prompt

智能系統研究所・共同指導：帥宏翰

電腦視覺、知識蒸餾、異質架構、提示微調、視覺轉換器、多層感知機、卷積神經網路、Computer vision、Knowledge distillation、Heterogeneous architecture、Prompt tuning、Vision transformer、Multi-layer perceptron layer、Convolution neural network

Development and Multi-Scenario Application Effect of an Intelligent Question-Answering System Based on In-Context Learning and Retrieval Augmented Generation Technology

智能系統研究所・共同指導：黃育綸

智能問答系統、GPT模型、自然語言處理、檢索增強生成（RAG）、上下文學習（In-context learning）、多場景實驗、系統設計、知識管理、Intelligent Question-Answering System、GPT Model、Natural Language Processing、Retrieval-Augmented Generation (RAG)、In-Context Learning、Multi-Scenario Experiments、System Design、Knowledge Management

A Gradient-Based One-Shot Model Channel Pruning Method for Object Detection Models

電子研究所・共同指導：簡鳳村

模型壓縮、模型剪枝、結構性剪枝、一次性剪枝、輕量化模型、物件偵測、Model Compression、Model Pruning、Structured Pruning、One-Shot Pruning、Light-Weight Model、Object Detection

Constructing Reality and Fantasy: Scene Creation via LLM-Enhanced Prompt Interpretation

電子研究所・共同指導：簡鳳村

文本生成圖像、佈局生成圖像、現實-幻想基準、擴散模型、大型語言模型(LLMs)、Text-to-Image Generation、Layout-to-Image Generation、Realistic-Fantasy Benchmark、Diffusion Model、Large Language Models (LLMs)

Using Vision Language Models to Assist Diffusion Models in Generating Realistic Interaction Scenes

智能系統研究所・共同指導：帥宏翰

多媒體圖像處理、文本生成圖像、擴散模型、視覺語言模型、人與物互動生成、Multimedia Image Processing、Text-to-Image Generation、Diffusion Model、Vision Language Models、Human-object Interaction Generation

Learning Efficient Interaction Anchor for HOI Detection

電機工程學系・共同指導：帥宏翰

人與物交互檢測（HOI）、交互錨點、Transformer、Human-object interaction detection (HOI)、Interaction anchor

Domain Adaptive Edge-guided DETR for Object Detection

電機資訊國際碩士學位學程・共同指導：帥宏翰

DETR、Domain Adaptation、Object Detection、Cityscapes

Scheduling-Sparsity Pruning Approach for Large Language Models

人工智慧技術與應用碩士學位學程・共同指導：帥宏翰

大型語言模型、模型壓縮、模型剪枝、稀疏性、large language model、model compression、model pruning、sparsity

Dual Memory-guided Probabilistic Model for Weakly-supervised Anomaly Detection

智能系統研究所・共同指導：帥宏翰

異常偵測、異常定位、擴散模型、擴散引導、弱監督式學習、記憶庫、anomaly detection、anomaly localization、diffusion model、diffusion guidance、weakly-supervised learning、memory bank

Color Splash Effects: using Semantic-Guided Referring Image Matting and Recoloring Network

電子研究所・共同指導：簡鳳村

色彩飛濺效果、視覺效果、影像處理、影像去背、影像重新著色、參考影像去背任務、Color Splash Effect、Visual Effects、Image Manipulation、Image Matting、Image Recoloring、Referring Image Matting task

Instance-level domain adaptation and pseudo label mining with Mean Teacher framework for Cross-domain Object Detection

電子研究所・共同指導：簡鳳村

深度學習、非監督領域適應、物件偵測、跨域物件偵測、Deep Learning、Unsupervise Domain Adaptation、Object Detection、Cross-domain Object Detection

Emotion Recognition based on Computer Vision

電子研究所・共同指導：簡鳳村

情緒識別、微表情辨識、群體情緒辨識、視覺語言模型、零樣本情緒辨識、emotion recognition、group affect recognition、vision-language model、zero-shot emotion recognition

Language-guided Negative Sample Mining for Open-vocabulary Object Detection

電子研究所・共同指導：簡鳳村

開放詞彙物件偵測、負樣本挖掘、視覺語言模型、標籤修正、open-vocabulary detection、negative sample mining、label revision

Sentiment Clustering and Learning to Prompt for Few-Shot Vision-Language Emotion Recognition

電子研究所・共同指導：簡鳳村

情緒辨識、跨模態任務、元學習、可學習提示、少樣本學習、零樣本學習、Emotion Recognition、Vision-Language Tasks、Meta-learning、Learnable Prompts、Few-shot Learning、Zero-shot Learning

From Machine Perception to Machine Cognition: The Deep Learning Based Visual Grounding and Human-Object Interaction Detection

電子研究所・共同指導：簡鳳村

深度學習、圖捲積、基於參考表達的視覺定位、人-物交互動作檢測、Deep Learning、Residual Graph Convolutional Networks、Visual Grounding、Human-Object Interaction Detection

Efficient Vehicle Detection and Tracking on Fisheye Traffic Monitoring Video in Pixel Domain and Compressed Domain

電機資訊國際學程・共同指導：杭學鳴

車輛檢測、車輛追蹤、魚眼相機、交通監控、壓縮域、vehicle detection、vehicle tracking、fisheye camera、traffic monitoring、compressed domain

Zero-shot Face-based Voice Conversion: Bottleneck-Free Speech Disentanglement with Adversarial Learning in the Real-world Scenario

人工智慧技術與應用碩士學位學程

多媒體應用、真實世界語音合成、去瓶頸語音轉換、跨域學習、臉與語音關係、對抗式學習、重參數化技巧、自注意力機制、Multimedia、Speech synthesis、Bottleneck-free voice conversion、Cross-modal learning、Face-voice relationship、Adversarial learning、Reparameterization trick、Mask self-attention mechanism

Semi-Supervised Pretraining and Dynamic Threshold for Effective Semi-Supervised Learning

電子研究所

深度學習、半監督學習、影像分類、一致性正則化、Deep Learning、Semi-Supervised Learning、Image Classification、Consistency Regularization

Improving Robustness of Facial Expression Recognition via Causal Learning

人工智慧技術與應用碩士學位學程

因果學習、面部表情辨識、偽關聯性、深度學習、穩健性、Causal Learning、Facial Expression Recognition、Spurious Correlation、Deep Learning、Robustness

Self-supervised Pre-training for Detection by Patch Rotation

電子研究所

自監督式學習、對比式學習、深度學習、物件偵測、卷積神經網路、Self-supervised Learning、Contrastive Learning、Deep Learning、Object Detection、Convolutional Neural Network

Trajectory prediction generative network via multi-attention in scene understanding and interaction parsing

人工智慧技術與應用碩士學位學程

路徑預測、自動駕駛、注意力機制、異構交通環境、遞歸神經網路、trajectory prediction、trajectory forecasting、autonomous driving、heterogeneous agents、recurrent neural network

Knowledge Distillation for Object Detectors via Attention Enhancement

電子研究所

卷積神經網絡、機器學習、物件偵測、模型壓縮、知識蒸餾、convolutional neural network、machine learning、object detection、model compression、knowledge distillation

Heterogeneous Federated Learning through Robust Multi­-branch Network

電子研究所

異質、聯邦學習、對抗式訓練、深度學習、神經網路、多分支網路、Heterogeneous、Federated Learning、Adversarial Training、Deep Learning、Neural Network、Multi­-branch Network

A Study of Practical Algorithms and Benchmarks for Network Anomaly Detection

人工智慧技術與應用碩士學位學程・共同指導：帥宏翰

網路異常偵測、入侵檢測資料集、演算法、機器學習、集成學習、神經網路、Network anomaly detection、Network-based intrusion detection dataset、Algorithm、Machine learning、Ensemble learning、Neural network

Semantic-guided vs. Co-attention Feature-remapping Virtual Try-on with Sequential Poses

電子研究所

虛擬試穿、語義分割、協同注意、Virtual Try-on、Semantic Segmentation、Co-attention

A Windshield Image Enhancer Based on a Two-way Generative Adversarial Network

電子研究所・共同指導：賴伯承

擋風玻璃影像強化器、生成對抗網路、感知損失函數、獨立正則化層、Windshield image enhancement、generative adversarial network、perceptual loss、individual instance normalization