[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"similar-ndb796--Deep-Learning-Paper-Review-and-Practice":3,"tool-ndb796--Deep-Learning-Paper-Review-and-Practice":64},[4,17,27,35,43,56],{"id":5,"name":6,"github_repo":7,"description_zh":8,"stars":9,"difficulty_score":10,"last_commit_at":11,"category_tags":12,"status":16},3808,"stable-diffusion-webui","AUTOMATIC1111\u002Fstable-diffusion-webui","stable-diffusion-webui 是一个基于 Gradio 构建的网页版操作界面，旨在让用户能够轻松地在本地运行和使用强大的 Stable Diffusion 图像生成模型。它解决了原始模型依赖命令行、操作门槛高且功能分散的痛点，将复杂的 AI 绘图流程整合进一个直观易用的图形化平台。\n\n无论是希望快速上手的普通创作者、需要精细控制画面细节的设计师，还是想要深入探索模型潜力的开发者与研究人员，都能从中获益。其核心亮点在于极高的功能丰富度：不仅支持文生图、图生图、局部重绘（Inpainting）和外绘（Outpainting）等基础模式，还独创了注意力机制调整、提示词矩阵、负向提示词以及“高清修复”等高级功能。此外，它内置了 GFPGAN 和 CodeFormer 等人脸修复工具，支持多种神经网络放大算法，并允许用户通过插件系统无限扩展能力。即使是显存有限的设备，stable-diffusion-webui 也提供了相应的优化选项，让高质量的 AI 艺术创作变得触手可及。",162132,3,"2026-04-05T11:01:52",[13,14,15],"开发框架","图像","Agent","ready",{"id":18,"name":19,"github_repo":20,"description_zh":21,"stars":22,"difficulty_score":23,"last_commit_at":24,"category_tags":25,"status":16},1381,"everything-claude-code","affaan-m\u002Feverything-claude-code","everything-claude-code 是一套专为 AI 编程助手（如 Claude Code、Codex、Cursor 等）打造的高性能优化系统。它不仅仅是一组配置文件，而是一个经过长期实战打磨的完整框架，旨在解决 AI 代理在实际开发中面临的效率低下、记忆丢失、安全隐患及缺乏持续学习能力等核心痛点。\n\n通过引入技能模块化、直觉增强、记忆持久化机制以及内置的安全扫描功能，everything-claude-code 能显著提升 AI 在复杂任务中的表现，帮助开发者构建更稳定、更智能的生产级 AI 代理。其独特的“研究优先”开发理念和针对 Token 消耗的优化策略，使得模型响应更快、成本更低，同时有效防御潜在的攻击向量。\n\n这套工具特别适合软件开发者、AI 研究人员以及希望深度定制 AI 工作流的技术团队使用。无论您是在构建大型代码库，还是需要 AI 协助进行安全审计与自动化测试，everything-claude-code 都能提供强大的底层支持。作为一个曾荣获 Anthropic 黑客大奖的开源项目，它融合了多语言支持与丰富的实战钩子（hooks），让 AI 真正成长为懂上",138956,2,"2026-04-05T11:33:21",[13,15,26],"语言模型",{"id":28,"name":29,"github_repo":30,"description_zh":31,"stars":32,"difficulty_score":23,"last_commit_at":33,"category_tags":34,"status":16},2271,"ComfyUI","Comfy-Org\u002FComfyUI","ComfyUI 是一款功能强大且高度模块化的视觉 AI 引擎，专为设计和执行复杂的 Stable Diffusion 图像生成流程而打造。它摒弃了传统的代码编写模式，采用直观的节点式流程图界面，让用户通过连接不同的功能模块即可构建个性化的生成管线。\n\n这一设计巧妙解决了高级 AI 绘图工作流配置复杂、灵活性不足的痛点。用户无需具备编程背景，也能自由组合模型、调整参数并实时预览效果，轻松实现从基础文生图到多步骤高清修复等各类复杂任务。ComfyUI 拥有极佳的兼容性，不仅支持 Windows、macOS 和 Linux 全平台，还广泛适配 NVIDIA、AMD、Intel 及苹果 Silicon 等多种硬件架构，并率先支持 SDXL、Flux、SD3 等前沿模型。\n\n无论是希望深入探索算法潜力的研究人员和开发者，还是追求极致创作自由度的设计师与资深 AI 绘画爱好者，ComfyUI 都能提供强大的支持。其独特的模块化架构允许社区不断扩展新功能，使其成为当前最灵活、生态最丰富的开源扩散模型工具之一，帮助用户将创意高效转化为现实。",107662,"2026-04-03T11:11:01",[13,14,15],{"id":36,"name":37,"github_repo":38,"description_zh":39,"stars":40,"difficulty_score":23,"last_commit_at":41,"category_tags":42,"status":16},3704,"NextChat","ChatGPTNextWeb\u002FNextChat","NextChat 是一款轻量且极速的 AI 助手，旨在为用户提供流畅、跨平台的大模型交互体验。它完美解决了用户在多设备间切换时难以保持对话连续性，以及面对众多 AI 模型不知如何统一管理的痛点。无论是日常办公、学习辅助还是创意激发，NextChat 都能让用户随时随地通过网页、iOS、Android、Windows、MacOS 或 Linux 端无缝接入智能服务。\n\n这款工具非常适合普通用户、学生、职场人士以及需要私有化部署的企业团队使用。对于开发者而言，它也提供了便捷的自托管方案，支持一键部署到 Vercel 或 Zeabur 等平台。\n\nNextChat 的核心亮点在于其广泛的模型兼容性，原生支持 Claude、DeepSeek、GPT-4 及 Gemini Pro 等主流大模型，让用户在一个界面即可自由切换不同 AI 能力。此外，它还率先支持 MCP（Model Context Protocol）协议，增强了上下文处理能力。针对企业用户，NextChat 提供专业版解决方案，具备品牌定制、细粒度权限控制、内部知识库整合及安全审计等功能，满足公司对数据隐私和个性化管理的高标准要求。",87618,"2026-04-05T07:20:52",[13,26],{"id":44,"name":45,"github_repo":46,"description_zh":47,"stars":48,"difficulty_score":23,"last_commit_at":49,"category_tags":50,"status":16},2268,"ML-For-Beginners","microsoft\u002FML-For-Beginners","ML-For-Beginners 是由微软推出的一套系统化机器学习入门课程，旨在帮助零基础用户轻松掌握经典机器学习知识。这套课程将学习路径规划为 12 周，包含 26 节精炼课程和 52 道配套测验，内容涵盖从基础概念到实际应用的完整流程，有效解决了初学者面对庞大知识体系时无从下手、缺乏结构化指导的痛点。\n\n无论是希望转型的开发者、需要补充算法背景的研究人员，还是对人工智能充满好奇的普通爱好者，都能从中受益。课程不仅提供了清晰的理论讲解，还强调动手实践，让用户在循序渐进中建立扎实的技能基础。其独特的亮点在于强大的多语言支持，通过自动化机制提供了包括简体中文在内的 50 多种语言版本，极大地降低了全球不同背景用户的学习门槛。此外，项目采用开源协作模式，社区活跃且内容持续更新，确保学习者能获取前沿且准确的技术资讯。如果你正寻找一条清晰、友好且专业的机器学习入门之路，ML-For-Beginners 将是理想的起点。",84991,"2026-04-05T10:45:23",[14,51,52,53,15,54,26,13,55],"数据工具","视频","插件","其他","音频",{"id":57,"name":58,"github_repo":59,"description_zh":60,"stars":61,"difficulty_score":10,"last_commit_at":62,"category_tags":63,"status":16},3128,"ragflow","infiniflow\u002Fragflow","RAGFlow 是一款领先的开源检索增强生成（RAG）引擎，旨在为大语言模型构建更精准、可靠的上下文层。它巧妙地将前沿的 RAG 技术与智能体（Agent）能力相结合，不仅支持从各类文档中高效提取知识，还能让模型基于这些知识进行逻辑推理和任务执行。\n\n在大模型应用中，幻觉问题和知识滞后是常见痛点。RAGFlow 通过深度解析复杂文档结构（如表格、图表及混合排版），显著提升了信息检索的准确度，从而有效减少模型“胡编乱造”的现象，确保回答既有据可依又具备时效性。其内置的智能体机制更进一步，使系统不仅能回答问题，还能自主规划步骤解决复杂问题。\n\n这款工具特别适合开发者、企业技术团队以及 AI 研究人员使用。无论是希望快速搭建私有知识库问答系统，还是致力于探索大模型在垂直领域落地的创新者，都能从中受益。RAGFlow 提供了可视化的工作流编排界面和灵活的 API 接口，既降低了非算法背景用户的上手门槛，也满足了专业开发者对系统深度定制的需求。作为基于 Apache 2.0 协议开源的项目，它正成为连接通用大模型与行业专有知识之间的重要桥梁。",77062,"2026-04-04T04:44:48",[15,14,13,26,54],{"id":65,"github_repo":66,"name":67,"description_en":68,"description_zh":69,"ai_summary_zh":69,"readme_en":70,"readme_zh":71,"quickstart_zh":72,"use_case_zh":73,"hero_image_url":74,"owner_login":75,"owner_name":76,"owner_avatar_url":77,"owner_bio":78,"owner_company":79,"owner_location":80,"owner_email":81,"owner_twitter":82,"owner_website":82,"owner_url":83,"languages":84,"stars":89,"forks":90,"last_commit_at":91,"license":82,"difficulty_score":23,"env_os":92,"env_gpu":92,"env_ram":92,"env_deps":93,"category_tags":96,"github_topics":97,"view_count":23,"oss_zip_url":82,"oss_zip_packed_at":82,"status":16,"created_at":101,"updated_at":102,"faqs":103,"releases":114},1320,"ndb796\u002FDeep-Learning-Paper-Review-and-Practice","Deep-Learning-Paper-Review-and-Practice","꼼꼼한 딥러닝 논문 리뷰와 코드 실습","Deep-Learning-Paper-Review-and-Practice 是一份“论文精读 + 代码实战”的深度学习学习地图。它把近年最常被引用的视觉、NLP、生成模型等方向的顶会论文，拆成通俗易懂的讲解视频、要点 PDF，并配套可直接运行的 Jupyter Notebook，让你读完就能动手复现。  \n它解决了“论文看不懂、代码跑不通、复现耗时长”的三大痛点：作者把晦涩公式翻译成中文笔记，把环境配置、超参数、训练技巧都提前调好，你只需一键运行即可看到结果。  \n适合想快速入门或深入的研究人员、算法工程师、高年级本科生，也适合需要案例教学的高校教师。  \n亮点在于“一条龙”体验：每篇论文都提供原论文链接、中文视频讲解、精炼 PDF、完整训练\u002F推理代码，甚至同一模型在 MNIST、CIFAR-10、ImageNet 等多数据集上的对比实验，真正做到“读一篇、会一篇、用一篇”。","### 꼼꼼한 딥러닝 논문 리뷰와 코드 실습: Deep Learning Paper Review and Practice\n\n* 꼼꼼한 딥러닝 논문 리뷰와 코드 실습을 위한 저장소입니다.\n* 최신 논문 위주로, 많은 인기를 끌고 있는 다양한 딥러닝 논문을 소개합니다.\n* 질문 사항은 본 저장소의 \u003Cb>[이슈(Issues)](https:\u002F\u002Fgithub.com\u002Fndb796\u002FDeep-Learning-Paper-Review-and-Practice\u002Fissues)\u003C\u002Fb> 탭에 남겨주세요.\n\n#### Image Recognition (이미지 인식)\n\n* End-to-End Object Detection with Transformers (ECCV 2020)\n    * [Original Paper Link](https:\u002F\u002Farxiv.org\u002Fabs\u002F2005.12872) \u002F [Paper Review Video](https:\u002F\u002Fwww.youtube.com\u002Fwatch?v=hCWUTvVrG7E) \u002F [Summary PDF](\u002Flecture_notes\u002FDETR.pdf) \u002F Code Practice\n* Searching for MobileNetV3 (ICCV 2019)\n    * [Original Paper Link](https:\u002F\u002Farxiv.org\u002Fabs\u002F1905.02244) \u002F Paper Review Video \u002F Summary PDF \u002F Code Practice\n* Deep Residual Learning for Image Recognition (CVPR 2016)\n    * [Original Paper Link](https:\u002F\u002Farxiv.org\u002Fabs\u002F1512.03385) \u002F [Paper Review Video](https:\u002F\u002Fwww.youtube.com\u002Fwatch?v=671BsKl8d0E) \u002F [Summary PDF](\u002Flecture_notes\u002FResNet.pdf) \u002F [MNIST](\u002Fcode_practices\u002FResNet18_MNIST_Train.ipynb) \u002F [CIFAR-10](\u002Fcode_practices\u002FResNet18_CIFAR10_Train.ipynb) \u002F [ImageNet](\u002Fcode_practices\u002FPretrained_ResNet18_ImageNet_Test.ipynb)\n* Arbitrary Style Transfer in Real-time with Adaptive Instance Normalization (ICCV 2017)\n    * [Original Paper Link](https:\u002F\u002Farxiv.org\u002Fabs\u002F1703.06868) \u002F [Paper Review Video](https:\u002F\u002Fwww.youtube.com\u002Fwatch?v=OM-6zYYRYfg) \u002F [Summary PDF](\u002Flecture_notes\u002FAdaIN_Style_Transfer.pdf) \u002F [Code Practice](\u002Fcode_practices\u002FAdaIN_Style_Transfer_Tutorial.ipynb)\n* Image Style Transfer Using Convolutional Neural Networks (CVPR 2016)\n    * [Original Paper Link](https:\u002F\u002Fwww.cv-foundation.org\u002Fopenaccess\u002Fcontent_cvpr_2016\u002Fpapers\u002FGatys_Image_Style_Transfer_CVPR_2016_paper.pdf) \u002F [Paper Review Video](https:\u002F\u002Fwww.youtube.com\u002Fwatch?v=va3e2c4uKJk) \u002F [Summary PDF](\u002Flecture_notes\u002FStyle%20Transfer.pdf) \u002F [Code Practice](\u002Fcode_practices\u002FStyle_Transfer_Tutorial.ipynb)\n* Faster R-CNN: Towards Real-Time Object Detection with Region Proposal Networks (NIPS 2015)\n    * [Original Paper Link](https:\u002F\u002Farxiv.org\u002Fabs\u002F1506.01497) \u002F Paper Review Video \u002F [Summary PDF](\u002Flecture_notes\u002FFaster_R-CNN.pdf) \u002F Code Practice\n\n#### Natural Language Processing (자연어 처리)\n\n* Single Headed Attention RNN: Stop Thinking With Your Head (2020)\n    * [Original Paper Link](https:\u002F\u002Farxiv.org\u002Fabs\u002F1911.11423) \u002F Paper Review Video \u002F Summary PDF \u002F Code Practice\n* BERT: Pre-training of Deep Bidirectional Transformers for Language Understanding (NAACL 2019)\n    * [Original Paper Link](https:\u002F\u002Farxiv.org\u002Fabs\u002F1810.04805) \u002F Paper Review Video \u002F Summary PDF \u002F Code Practice\n* Attention is All You Need (NIPS 2017)\n    * [Original Paper Link](https:\u002F\u002Farxiv.org\u002Fabs\u002F1706.03762) \u002F [Paper Review Video](https:\u002F\u002Fwww.youtube.com\u002Fwatch?v=AA621UofTUA) \u002F [Summary PDF](\u002Flecture_notes\u002FTransformer.pdf) \u002F [Code Practice](\u002Fcode_practices\u002FAttention_is_All_You_Need_Tutorial_(German_English).ipynb)\n* Neural Machine Translation by Jointly Learning to Align and Translate (ICLR 2015 Oral)\n    * [Original Paper Link](https:\u002F\u002Farxiv.org\u002Fabs\u002F1409.0473) \u002F Paper Review Video \u002F Summary PDF \u002F [Code Practice](\u002Fcode_practices\u002FSequence_to_Sequence_with_Attention_Tutorial.ipynb)\n* Show and Tell: A Neural Image Caption Generator (CVPR 2015)\n    * [Original Paper Link](https:\u002F\u002Farxiv.org\u002Fabs\u002F1411.4555) \u002F Paper Review Video \u002F Summary PDF \u002F [Code Practice](\u002Fcode_practices\u002FNeural_Image_Captioning_(NIC)_Using_ResNet_101.ipynb)\n* Sequence to Sequence Learning with Neural Networks (NIPS 2014)\n    * [Original Paper Link](https:\u002F\u002Farxiv.org\u002Fabs\u002F1409.3215) \u002F [Paper Review Video](https:\u002F\u002Fwww.youtube.com\u002Fwatch?v=4DzKM0vgG1Y) \u002F [Summary PDF](\u002Flecture_notes\u002FSeq2Seq.pdf) \u002F [Code Practice](\u002Fcode_practices\u002FSequence_to_Sequence_with_LSTM_Tutorial.ipynb)\n\n#### Generative Model & Super Resolution (생성 모델 & 해상도 복원)\n\n* Meta-Transfer Learning for Zero-Shot Super-Resolution (CVPR 2020)\n    * [Original Paper Link](https:\u002F\u002Farxiv.org\u002Fabs\u002F2002.12213) \u002F [Paper Review Video](https:\u002F\u002Fwww.youtube.com\u002Fwatch?v=PUtFz4vqXHQ) \u002F [Summary PDF](\u002Flecture_notes\u002FMZSR.pdf) \u002F Code Practice\n* SinGAN: Learning a Generative Model from a Single Natural Image (ICCV 2019)\n    * [Original Paper Link](https:\u002F\u002Farxiv.org\u002Fabs\u002F1905.01164) \u002F Paper Review Video \u002F Summary PDF \u002F Code Practice\n* A Style-Based Generator Architecture for Generative Adversarial Networks (CVPR 2019)\n    * [Original Paper Link](https:\u002F\u002Farxiv.org\u002Fabs\u002F1812.04948) \u002F Paper Review Video \u002F [Summary PDF](\u002Flecture_notes\u002FStyleGAN.pdf) \u002F Code Practice\n* StarGAN: Unified Generative Adversarial Networks for Multi-Domain Image-to-Image Translation (CVPR 2018 Oral)\n    * [Original Paper Link](https:\u002F\u002Farxiv.org\u002Fabs\u002F1711.09020) \u002F [Paper Review Video](https:\u002F\u002Fwww.youtube.com\u002Fwatch?v=-r9M4Cj9o_8) \u002F [Summary PDF](\u002Flecture_notes\u002FStarGAN.pdf) \u002F [Code Practice](\u002Fcode_practices\u002FStarGAN_Tutorial.ipynb)\n* Image-to-Image Translation with Conditional Adversarial Networks (CVPR 2017)\n    * [Original Paper Link](https:\u002F\u002Farxiv.org\u002Fabs\u002F1611.07004) \u002F [Paper Review Video](https:\u002F\u002Fwww.youtube.com\u002Fwatch?v=ImiD4npRj7k) \u002F [Summary PDF](\u002Flecture_notes\u002FPix2Pix.pdf) \u002F [Code Practice](\u002Fcode_practices\u002FPix2Pix_for_Facades.ipynb)\n* Generative Adversarial Nets (NIPS 2014)\n    * [Original Paper Link](https:\u002F\u002Farxiv.org\u002Fabs\u002F1406.2661) \u002F [Paper Review Video](https:\u002F\u002Fwww.youtube.com\u002Fwatch?v=AVvlDmhHgC4) \u002F [Summary PDF](\u002Flecture_notes\u002FGAN.pdf) \u002F [Code Practice](\u002Fcode_practices\u002FGAN_for_MNIST_Tutorial.ipynb)\n\n#### Modeling & Optimization (모델링 & 최적화)\n\n* Bag of Tricks for Image Classification (CVPR 2019)\n    * [Original Paper Link](https:\u002F\u002Farxiv.org\u002Fabs\u002F1812.01187) \u002F Paper Review Video \u002F [Summary PDF](\u002Flecture_notes\u002FBag_of_Tricks_for_Image_Classification.pdf)\n    * [CIFAR-10](\u002Fcode_practices\u002FResNet18_CIFAR10_Basic_Training.ipynb) \u002F [CIFAR-10 with Label Smoothing](\u002Fcode_practices\u002FResNet18_CIFAR10_Training_with_Label_Smoothing.ipynb) \u002F [CIFAR-10 with Input Mixup](\u002Fcode_practices\u002FResNet18_CIFAR10_Training_with_Input_Mixup.ipynb) \u002F [CIFAR-10 with Label Smoothing and Input Mixup](\u002Fcode_practices\u002FResNet18_CIFAR10_Training_with_Input_Mixup_and_Label_Smoothing.ipynb)\n* Deep Compression: Compressing Deep Neural Networks with Pruning, Trained Quantization and Huffman Coding (ICLR 2016 Oral)\n    * [Original Paper Link](https:\u002F\u002Farxiv.org\u002Fabs\u002F1510.00149) \u002F Paper Review Video \u002F Summary PDF \u002F Code Practice\n* Batch normalization: Accelerating deep network training by reducing internal covariate shift (PMLR 2015)\n    * [Original Paper Link](https:\u002F\u002Farxiv.org\u002Fabs\u002F1502.03167) \u002F [Paper Review Video](https:\u002F\u002Fwww.youtube.com\u002Fwatch?v=58fuWVu5DVU) \u002F [Summary PDF](\u002Flecture_notes\u002FBatch_Normalization.pdf) \u002F [Code Practice](\u002Fcode_practices\u002FBatch_Normalization_Evaluation_(with_Residual_Connection).ipynb)\n\n#### Adversarial Examples & Backdoor Attacks (적대적 예제 & 백도어 공격)\n\n* HopSkipJumpAttack: A Query-Efficient Decision-Based Attack (S&P 2020)\n    * [Original Paper Link](https:\u002F\u002Farxiv.org\u002Fabs\u002F1904.02144) \u002F Paper Review Video\u002F Summary PDF \u002F [Targeted Attack](\u002Fcode_practices\u002FTargeted_HopSkipJumpAttack_Using_CIFAR10.ipynb) \u002F [Untargeted Attack](\u002Fcode_practices\u002FUntargeted_HopSkipJumpAttack_Using_CIFAR10.ipynb)\n* Breaking certified defenses: Semantic adversarial examples with spoofed robustness certificates (ICLR 2020)\n    * [Original Paper Link](https:\u002F\u002Farxiv.org\u002Fabs\u002F2003.08937) \u002F [Paper Review Video](https:\u002F\u002Fwww.youtube.com\u002Fwatch?v=D1j3QiXPRag) \u002F [Summary PDF](\u002Flecture_notes\u002FShadow_Attack.pdf) \u002F [Code Practice](\u002Fcode_practices\u002FShadow_Attack_Tutorial.ipynb)\n* Sign-OPT: A Query-Efficient Hard-label Adversarial Attack (ICLR 2020)\n    * [Original Paper Link](https:\u002F\u002Farxiv.org\u002Fabs\u002F1909.10773) \u002F Paper Review Video \u002F Summary PDF \u002F [MNIST](\u002Fcode_practices\u002FSign_OPT_Attack_for_MNIST.ipynb) \u002F [CIFAR-10](\u002Fcode_practices\u002FSign_OPT_Attack_for_CIFAR_10.ipynb)\n* Is BERT Really Robust? A Strong Baseline for Natural Language Attack on Text Classification and Entailment (AAAI 2020 Oral)\n    * [Original Paper Link](https:\u002F\u002Farxiv.org\u002Fabs\u002F1907.11932) \u002F [Paper Review Video](https:\u002F\u002Fwww.youtube.com\u002Fwatch?v=EF-IYFTKZiE) \u002F [Summary PDF](\u002Flecture_notes\u002FTextFooler.pdf) \u002F [Code Practice](\u002Fcode_practices\u002FTextFooler_Tutorial.ipynb)\n* Query-Efficient Hard-label Black-box Attack: An Optimization-based Approach (ICLR 2019)\n    * [Original Paper Link](https:\u002F\u002Farxiv.org\u002Fabs\u002F1807.04457) \u002F Paper Review Video \u002F [Summary PDF](\u002Flecture_notes\u002FOPT_Attack.pdf) \u002F [MNIST](\u002Fcode_practices\u002FOpt_Attack_for_MNIST.ipynb) \u002F [CIFAR-10](\u002Fcode_practices\u002FOpt_Attack_for_CIFAR_10.ipynb)\n* Boosting Adversarial Attacks with Momentum (CVPR 2018 Spotlight)\n    * [Original Paper Link](https:\u002F\u002Farxiv.org\u002Fabs\u002F1710.06081) \u002F Paper Review Video \u002F [Summary PDF](\u002Flecture_notes\u002FBoosting_Adversarial_Attacks_with_Momentum.pdf) \u002F [CIFAR-10](\u002Fcode_practices\u002FMI_FGSM_Attack_for_CIFAR_10.ipynb) \u002F [ImageNet](\u002Fcode_practices\u002FMI_FGSM_Attack_for_ImageNet.ipynb)\n* Poison Frogs! Targeted Clean-Label Poisoning Attacks on Neural Networks (NIPS 2018)\n    * [Original Paper Link](https:\u002F\u002Farxiv.org\u002Fabs\u002F1804.00792) \u002F Paper Review Video \u002F [Summary PDF](\u002Flecture_notes\u002FPoison_Frogs.pdf) \u002F [ResNet](\u002Fcode_practices\u002FOne_Shot_Kill_Poison_Attack_ResNet.ipynb) \u002F [AlexNet](\u002Fcode_practices\u002FOne_Shot_Kill_Poison_Attack_AlexNet.ipynb)\n* Decision-Based Adversarial Attacks: Reliable Attacks Against Black-Box Machine Learning Models (ICLR 2018)\n    * [Original Paper Link](https:\u002F\u002Farxiv.org\u002Fabs\u002F1712.04248) \u002F [Paper Review Video](https:\u002F\u002Fwww.youtube.com\u002Fwatch?v=3dX_SsO2mis) \u002F [Summary PDF](\u002Flecture_notes\u002FBoundary_Attack.pdf) \u002F Code Practice\n\n### 지난 논문 리뷰 콘텐츠\n\n* Explaining and Harnessing Adversarial Examples (ICLR 2015)\n    * [Original Paper Link](https:\u002F\u002Farxiv.org\u002Fabs\u002F1412.6572) \u002F [Paper Review Video](https:\u002F\u002Fwww.youtube.com\u002Fwatch?v=99uxhAjNwps)\n* Towards Evaluating the Robustness of Neural Networks (S&P 2017)\n    * [Original Paper Link](https:\u002F\u002Farxiv.org\u002Fabs\u002F1608.04644) \u002F [Paper Review Video](https:\u002F\u002Fwww.youtube.com\u002Fwatch?v=9kRWHKPyfwQ)\n* Towards Deep Learning Models Resistant to Adversarial Attacks (ICLR 2018)\n    * [Original Paper Link](https:\u002F\u002Farxiv.org\u002Fabs\u002F1706.06083) \u002F [Paper Review Video](https:\u002F\u002Fwww.youtube.com\u002Fwatch?v=6RBpdAC9nwY)\n* Adversarial Examples Are Not Bugs, They Are Features (NIPS 2019)\n    * [Original Paper Link](https:\u002F\u002Farxiv.org\u002Fabs\u002F1905.02175) \u002F [Paper Review Video](https:\u002F\u002Fwww.youtube.com\u002Fwatch?v=Y7O47Kq8pmU)\n* Certified Robustness to Adversarial Examples with Differential Privacy (S&P 2019)\n    * [Original Paper Link](https:\u002F\u002Farxiv.org\u002Fabs\u002F1802.03471) \u002F [Paper Review Video](https:\u002F\u002Fwww.youtube.com\u002Fwatch?v=ySJUlEVlXfk)\n* Obfuscated Gradients Give a False Sense of Security (ICML 2018)\n    * [Original Paper Link](https:\u002F\u002Farxiv.org\u002Fabs\u002F1802.00420) \u002F [Paper Review Video](https:\u002F\u002Fwww.youtube.com\u002Fwatch?v=0O_Bxln9bTw)\n* Constructing Unrestricted Adversarial Examples with Generative Models (NIPS 2018)\n    * [Original Paper Link](https:\u002F\u002Farxiv.org\u002Fabs\u002F1805.07894) \u002F [Paper Review Video](https:\u002F\u002Fwww.youtube.com\u002Fwatch?v=IDtaVjJoV4g)\n* Adversarial Patch (NIPS 2018)\n    * [Original Paper Link](https:\u002F\u002Farxiv.org\u002Fabs\u002F1712.09665) \u002F [Paper Review Video](https:\u002F\u002Fwww.youtube.com\u002Fwatch?v=pOlPlTCfCQE)\n","### 详尽的深度学习论文综述与代码实践：Deep Learning Paper Review and Practice\n\n* 这是一个用于详尽的深度学习论文综述与代码实践的仓库。\n* 主要聚焦于最新论文，介绍多篇广受关注的深度学习研究。\n* 如有任何问题，请在本仓库的\u003Cb>[Issues](https:\u002F\u002Fgithub.com\u002Fndb796\u002FDeep-Learning-Paper-Review-and-Practice\u002Fissues)\u003C\u002Fb>标签页中留言。\n\n#### 图像识别\n\n* 使用Transformer实现端到端目标检测（ECCV 2020）\n    * [原文链接](https:\u002F\u002Farxiv.org\u002Fabs\u002F2005.12872) \u002F [论文综述视频](https:\u002F\u002Fwww.youtube.com\u002Fwatch?v=hCWUTvVrG7E) \u002F [摘要PDF](\u002Flecture_notes\u002FDETR.pdf) \u002F 代码实践\n* 寻找MobileNetV3（ICCV 2019）\n    * [原文链接](https:\u002F\u002Farxiv.org\u002Fabs\u002F1905.02244) \u002F 论文综述视频 \u002F 摘要PDF \u002F 代码实践\n* 用于图像识别的深度残差学习（CVPR 2016）\n    * [原文链接](https:\u002F\u002Farxiv.org\u002Fabs\u002F1512.03385) \u002F [论文综述视频](https:\u002F\u002Fwww.youtube.com\u002Fwatch?v=671BsKl8d0E) \u002F [摘要PDF](\u002Flecture_notes\u002FResNet.pdf) \u002F [MNIST](\u002Fcode_practices\u002FResNet18_MNIST_Train.ipynb) \u002F [CIFAR-10](\u002Fcode_practices\u002FResNet18_CIFAR10_Train.ipynb) \u002F [ImageNet](\u002Fcode_practices\u002FPretrained_ResNet18_ImageNet_Test.ipynb)\n* 基于自适应实例归一化的实时任意风格迁移（ICCV 2017）\n    * [原文链接](https:\u002F\u002Farxiv.org\u002Fabs\u002F1703.06868) \u002F [论文综述视频](https:\u002F\u002Fwww.youtube.com\u002Fwatch?v=OM-6zYYRYfg) \u002F [摘要PDF](\u002Flecture_notes\u002FAdaIN_Style_Transfer.pdf) \u002F [代码实践](\u002Fcode_practices\u002FAdaIN_Style_Transfer_Tutorial.ipynb)\n* 利用卷积神经网络进行图像风格迁移（CVPR 2016）\n    * [原文链接](https:\u002F\u002Fwww.cv-foundation.org\u002Fopenaccess\u002Fcontent_cvpr_2016\u002Fpapers\u002FGatys_Image_Style_Transfer_CVPR_2016_paper.pdf) \u002F [论文综述视频](https:\u002F\u002Fwww.youtube.com\u002Fwatch?v=va3e2c4uKJk) \u002F [摘要PDF](\u002Flecture_notes\u002FStyle%20Transfer.pdf) \u002F [代码实践](\u002Fcode_practices\u002FStyle_Transfer_Tutorial.ipynb)\n* Faster R-CNN：迈向基于区域建议网络的实时目标检测（NIPS 2015）\n    * [原文链接](https:\u002F\u002Farxiv.org\u002Fabs\u002F1506.01497) \u002F 论文综述视频 \u002F [摘要PDF](\u002Flecture_notes\u002FFaster_R-CNN.pdf) \u002F 代码实践\n\n#### 自然语言处理\n\n* 单头注意力RNN：停止用大脑思考（2020）\n    * [原文链接](https:\u002F\u002Farxiv.org\u002Fabs\u002F1911.11423) \u002F 论文综述视频 \u002F 摘要PDF \u002F 代码实践\n* BERT：用于语言理解的深度双向Transformer预训练（NAACL 2019）\n    * [原文链接](https:\u002F\u002Farxiv.org\u002Fabs\u002F1810.04805) \u002F 论文综述视频 \u002F 摘要PDF \u002F 代码实践\n* 注意力就是一切（NIPS 2017）\n    * [原文链接](https:\u002F\u002Farxiv.org\u002Fabs\u002F1706.03762) \u002F [论文综述视频](https:\u002F\u002Fwww.youtube.com\u002Fwatch?v=AA621UofTUA) \u002F [摘要PDF](\u002Flecture_notes\u002FTransformer.pdf) \u002F [代码实践](\u002Fcode_practices\u002FAttention_is_All_You_Need_Tutorial_(German_English).ipynb)\n* 通过联合学习对齐与翻译实现神经机器翻译（ICLR 2015口头报告）\n    * [原文链接](https:\u002F\u002Farxiv.org\u002Fabs\u002F1409.0473) \u002F 论文综述视频 \u002F 摘要PDF \u002F [代码实践](\u002Fcode_practices\u002FSequence_to_Sequence_with_Attention_Tutorial.ipynb)\n* 展示与讲述：神经图像字幕生成器（CVPR 2015）\n    * [原文链接](https:\u002F\u002Farxiv.org\u002Fabs\u002F1411.4555) \u002F 论文综述视频 \u002F 摘要PDF \u002F [代码实践](\u002Fcode_practices\u002FNeural_Image_Captioning_(NIC)_Using_ResNet_101.ipynb)\n* 使用神经网络的序列到序列学习（NIPS 2014）\n    * [原文链接](https:\u002F\u002Farxiv.org\u002Fabs\u002F1409.3215) \u002F [论文综述视频](https:\u002F\u002Fwww.youtube.com\u002Fwatch?v=4DzKM0vgG1Y) \u002F [摘要PDF](\u002Flecture_notes\u002FSeq2Seq.pdf) \u002F [代码实践](\u002Fcode_practices\u002FSequence_to_Sequence_with_LSTM_Tutorial.ipynb)\n\n#### 生成模型与超分辨率\n\n* 零样本超分辨率的元迁移学习（CVPR 2020）\n    * [原文链接](https:\u002F\u002Farxiv.org\u002Fabs\u002F2002.12213) \u002F [论文综述视频](https:\u002F\u002Fwww.youtube.com\u002Fwatch?v=PUtFz4vqXHQ) \u002F [摘要PDF](\u002Flecture_notes\u002FMZSR.pdf) \u002F 代码实践\n* SinGAN：从单张自然图像中学习生成模型（ICCV 2019）\n    * [原文链接](https:\u002F\u002Farxiv.org\u002Fabs\u002F1905.01164) \u002F 论文综述视频 \u002F 摘要PDF \u002F 代码实践\n* 基于风格的生成器架构用于生成对抗网络（CVPR 2019）\n    * [原文链接](https:\u002F\u002Farxiv.org\u002Fabs\u002F1812.04948) \u002F 论文综述视频 \u002F [摘要PDF](\u002Flecture_notes\u002FStyleGAN.pdf) \u002F 代码实践\n* StarGAN：用于多领域图像到图像转换的统一生成对抗网络（CVPR 2018口头报告）\n    * [原文链接](https:\u002F\u002Farxiv.org\u002Fabs\u002F1711.09020) \u002F [论文综述视频](https:\u002F\u002Fwww.youtube.com\u002Fwatch?v=-r9M4Cj9o_8) \u002F [摘要PDF](\u002Flecture_notes\u002FStarGAN.pdf) \u002F [代码实践](\u002Fcode_practices\u002FStarGAN_Tutorial.ipynb)\n* 带有条件对抗网络的图像到图像转换（CVPR 2017）\n    * [原文链接](https:\u002F\u002Farxiv.org\u002Fabs\u002F1611.07004) \u002F [论文综述视频](https:\u002F\u002Fwww.youtube.com\u002Fwatch?v=ImiD4npRj7k) \u002F [摘要PDF](\u002Flecture_notes\u002FPix2Pix.pdf) \u002F [代码实践](\u002Fcode_practices\u002FPix2Pix_for_Facades.ipynb)\n* 生成对抗网络（NIPS 2014）\n    * [原文链接](https:\u002F\u002Farxiv.org\u002Fabs\u002F1406.2661) \u002F [论文综述视频](https:\u002F\u002Fwww.youtube.com\u002Fwatch?v=AVvlDmhHgC4) \u002F [摘要PDF](\u002Flecture_notes\u002FGAN.pdf) \u002F [代码实践](\u002Fcode_practices\u002FGAN_for_MNIST_Tutorial.ipynb)\n\n#### 建模与优化\n\n* 图像分类的技巧大全（CVPR 2019）\n    * [原文链接](https:\u002F\u002Farxiv.org\u002Fabs\u002F1812.01187) \u002F 论文综述视频 \u002F [摘要PDF](\u002Flecture_notes\u002FBag_of_Tricks_for_Image_Classification.pdf)\n    * [CIFAR-10](\u002Fcode_practices\u002FResNet18_CIFAR10_Basic_Training.ipynb) \u002F [CIFAR-10带标签平滑](\u002Fcode_practices\u002FResNet18_CIFAR10_Training_with_Label_Smoothing.ipynb) \u002F [CIFAR-10带输入混洗](\u002Fcode_practices\u002FResNet18_CIFAR10_Training_with_Input_Mixup.ipynb) \u002F [CIFAR-10带标签平滑和输入混洗](\u002Fcode_practices\u002FResNet18_CIFAR10_Training_with_Input_Mixup_and_Label_Smoothing.ipynb)\n* 深度压缩：通过剪枝、训练量化和霍夫曼编码压缩深度神经网络（ICLR 2016口头报告）\n    * [原文链接](https:\u002F\u002Farxiv.org\u002Fabs\u002F1510.00149) \u002F 论文综述视频 \u002F 摘要PDF \u002F 代码实践\n* 批量归一化：通过减少内部协变量偏移加速深度网络训练（PMLR 2015）\n    * [原文链接](https:\u002F\u002Farxiv.org\u002Fabs\u002F1502.03167) \u002F [论文综述视频](https:\u002F\u002Fwww.youtube.com\u002Fwatch?v=58fuWVu5DVU) \u002F [摘要PDF](\u002Flecture_notes\u002FBatch_Normalization.pdf) \u002F [代码实践](\u002Fcode_practices\u002FBatch_Normalization_Evaluation_(with_Residual_Connection).ipynb)\n\n#### 对抗性样本与后门攻击\n\n* HopSkipJumpAttack：一种查询高效的基于决策的攻击（S&P 2020）\n    * [原文链接](https:\u002F\u002Farxiv.org\u002Fabs\u002F1904.02144) \u002F 论文综述视频 \u002F 摘要PDF \u002F [定向攻击](\u002Fcode_practices\u002FTargeted_HopSkipJumpAttack_Using_CIFAR10.ipynb) \u002F [非定向攻击](\u002Fcode_practices\u002FUntargeted_HopSkipJumpAttack_Using_CIFAR10.ipynb)\n* 突破认证防御：带有伪造鲁棒性证书的语义对抗样本（ICLR 2020）\n    * [原文链接](https:\u002F\u002Farxiv.org\u002Fabs\u002F2003.08937) \u002F [论文综述视频](https:\u002F\u002Fwww.youtube.com\u002Fwatch?v=D1j3QiXPRag) \u002F [摘要PDF](\u002Flecture_notes\u002FShadow_Attack.pdf) \u002F [代码实践](\u002Fcode_practices\u002FShadow_Attack_Tutorial.ipynb)\n* Sign-OPT：一种查询高效的硬标签对抗攻击（ICLR 2020）\n    * [原文链接](https:\u002F\u002Farxiv.org\u002Fabs\u002F1909.10773) \u002F 论文综述视频 \u002F 摘要PDF \u002F [MNIST](\u002Fcode_practices\u002FSign_OPT_Attack_for_MNIST.ipynb) \u002F [CIFAR-10](\u002Fcode_practices\u002FSign_OPT_Attack_for_CIFAR_10.ipynb)\n* BERT真的鲁棒吗？针对文本分类与蕴含任务的自然语言攻击强基线（AAAI 2020 口头报告）\n    * [原文链接](https:\u002F\u002Farxiv.org\u002Fabs\u002F1907.11932) \u002F [论文综述视频](https:\u002F\u002Fwww.youtube.com\u002Fwatch?v=EF-IYFTKZiE) \u002F [摘要PDF](\u002Flecture_notes\u002FTextFooler.pdf) \u002F [代码实践](\u002Fcode_practices\u002FTextFooler_Tutorial.ipynb)\n* 查询高效的硬标签黑盒攻击：一种基于优化的方法（ICLR 2019）\n    * [原文链接](https:\u002F\u002Farxiv.org\u002Fabs\u002F1807.04457) \u002F 论文综述视频 \u002F [摘要PDF](\u002Flecture_notes\u002FOPT_Attack.pdf) \u002F [MNIST](\u002Fcode_practices\u002FOpt_Attack_for_MNIST.ipynb) \u002F [CIFAR-10](\u002Fcode_practices\u002FOpt_Attack_for_CIFAR_10.ipynb)\n* 利用动量增强对抗攻击（CVPR 2018 焦点论文）\n    * [原文链接](https:\u002F\u002Farxiv.org\u002Fabs\u002F1710.06081) \u002F 论文综述视频 \u002F [摘要PDF](\u002Flecture_notes\u002FBoosting_Adversarial_Attacks_with_Momentum.pdf) \u002F [CIFAR-10](\u002Fcode_practices\u002FMI_FGSM_Attack_for_CIFAR_10.ipynb) \u002F [ImageNet](\u002Fcode_practices\u002FMI_FGSM_Attack_for_ImageNet.ipynb)\n* 毒蛙！针对神经网络的定向纯标签投毒攻击（NIPS 2018）\n    * [原文链接](https:\u002F\u002Farxiv.org\u002Fabs\u002F1804.00792) \u002F 论文综述视频 \u002F [摘要PDF](\u002Flecture_notes\u002FPoison_Frogs.pdf) \u002F [ResNet](\u002Fcode_practices\u002FOne_Shot_Kill_Poison_Attack_ResNet.ipynb) \u002F [AlexNet](\u002Fcode_practices\u002FOne_Shot_Kill_Poison_Attack_AlexNet.ipynb)\n* 基于决策的对抗攻击：针对黑盒机器学习模型的可靠攻击（ICLR 2018）\n    * [原文链接](https:\u002F\u002Farxiv.org\u002Fabs\u002F1712.04248) \u002F [论文综述视频](https:\u002F\u002Fwww.youtube.com\u002Fwatch?v=3dX_SsO2mis) \u002F [摘要PDF](\u002Flecture_notes\u002FBoundary_Attack.pdf) \u002F 代码实践\n\n\n\n### 过去论文综述内容\n\n* 解释并利用对抗样本（ICLR 2015）\n    * [原文链接](https:\u002F\u002Farxiv.org\u002Fabs\u002F1412.6572) \u002F [论文综述视频](https:\u002F\u002Fwww.youtube.com\u002Fwatch?v=99uxhAjNwps)\n* 探索评估神经网络的鲁棒性（S&P 2017）\n    * [原文链接](https:\u002F\u002Farxiv.org\u002Fabs\u002F1608.04644) \u002F [论文综述视频](https:\u002F\u002Fwww.youtube.com\u002Fwatch?v=9kRWHKPyfwQ)\n* 构建对对抗攻击具有抵抗力的深度学习模型（ICLR 2018）\n    * [原文链接](https:\u002F\u002Farxiv.org\u002Fabs\u002F1706.06083) \u002F [论文综述视频](https:\u002F\u002Fwww.youtube.com\u002Fwatch?v=6RBpdAC9nwY)\n* 对抗样本不是漏洞，而是特征（NIPS 2019）\n    * [原文链接](https:\u002F\u002Farxiv.org\u002Fabs\u002F1905.02175) \u002F [论文综述视频](https:\u002F\u002Fwww.youtube.com\u002Fwatch?v=Y7O47Kq8pmU)\n* 带有差分隐私的对抗样本认证鲁棒性（S&P 2019）\n    * [原文链接](https:\u002F\u002Farxiv.org\u002Fabs\u002F1802.03471) \u002F [论文综述视频](https:\u002F\u002Fwww.youtube.com\u002Fwatch?v=ySJUlEVlXfk)\n* 混淆梯度会带来虚假的安全感（ICML 2018）\n    * [原文链接](https:\u002F\u002Farxiv.org\u002Fabs\u002F1802.00420) \u002F [论文综述视频](https:\u002F\u002Fwww.youtube.com\u002Fwatch?v=0O_Bxln9bTw)\n* 使用生成模型构造无限制的对抗样本（NIPS 2018）\n    * [原文链接](https:\u002F\u002Farxiv.org\u002Fabs\u002F1805.07894) \u002F [论文综述视频](https:\u002F\u002Fwww.youtube.com\u002Fwatch?v=IDtaVjJoV4g)\n* 对抗补丁（NIPS 2018）\n    * [原文链接](https:\u002F\u002Farxiv.org\u002Fabs\u002F1712.09665) \u002F [论文综述视频](https:\u002F\u002Fwww.youtube.com\u002Fwatch?v=pOlPlTCfCQE)","# Deep-Learning-Paper-Review-and-Practice 中文快速上手指南\n\n## 环境准备\n\n- **操作系统**：Linux \u002F macOS \u002F Windows 10+  \n- **Python**：≥ 3.7（推荐 3.8+）  \n- **硬件**：  \n  - CPU 可运行，GPU（CUDA 11.x + cuDNN 8.x）可加速  \n- **依赖**：  \n  - PyTorch ≥ 1.9  \n  - TensorFlow ≥ 2.6（可选，部分 notebook 用到）  \n  - JupyterLab \u002F VS Code + Jupyter 插件  \n  - Git ≥ 2.20  \n\n## 安装步骤\n\n1. 克隆仓库（国内镜像加速）  \n   ```bash\n   git clone https:\u002F\u002Fghproxy.com\u002Fhttps:\u002F\u002Fgithub.com\u002Fndb796\u002FDeep-Learning-Paper-Review-and-Practice.git\n   cd Deep-Learning-Paper-Review-and-Practice\n   ```\n\n2. 创建并激活虚拟环境  \n   ```bash\n   python -m venv venv\n   source venv\u002Fbin\u002Factivate          # Linux\u002FmacOS\n   venv\\Scripts\\activate             # Windows\n   ```\n\n3. 一键安装依赖  \n   ```bash\n   pip install -r requirements.txt -i https:\u002F\u002Fpypi.tuna.tsinghua.edu.cn\u002Fsimple\n   ```\n\n   若仓库未提供 `requirements.txt`，可手动安装核心包：  \n   ```bash\n   pip install torch torchvision torchaudio --extra-index-url https:\u002F\u002Fdownload.pytorch.org\u002Fwhl\u002Fcu118\n   pip install jupyter matplotlib seaborn tqdm opencv-python -i https:\u002F\u002Fpypi.tuna.tsinghua.edu.cn\u002Fsimple\n   ```\n\n## 基本使用\n\n1. 启动 JupyterLab  \n   ```bash\n   jupyter lab\n   ```\n   浏览器自动打开 `http:\u002F\u002Flocalhost:8888`。\n\n2. 运行示例：ResNet18 在 CIFAR-10 训练  \n   - 在左侧导航栏进入  \n     ```\n     code_practices\u002FResNet18_CIFAR10_Train.ipynb\n     ```\n   - 按顺序执行所有 Cell（Shift+Enter）即可开始训练。  \n   - 默认 10 个 epoch，GPU 上约 5 分钟完成。\n\n3. 查看结果  \n   - 训练日志实时输出在 notebook 中。  \n   - 生成的模型权重保存在同级目录 `*.pth`。  \n\n至此，你已跑通第一个示例，可继续探索其他 notebook 或阅读 `\u002Flecture_notes\u002F` 中的中文\u002F英文总结 PDF。","某高校计算机系研二学生小赵，正在做“基于 Transformer 的实时风格迁移”课题，需要在 4 周内复现并改进一篇 ICCV 论文，作为中期答辩的核心成果。\n\n### 没有 Deep-Learning-Paper-Review-and-Practice 时\n- 先花 2 天在 arXiv 上搜到 20 篇相关论文，却分不清哪篇是 ICCV 2017 AdaIN 还是后续改进，阅读顺序混乱。  \n- 啃原论文时，公式与伪代码夹杂，缺乏直观讲解，3 天过去仍对“Adaptive Instance Normalization”细节一知半解。  \n- 从零搭 PyTorch 框架：调 dataloader、写 loss、配环境，踩坑 CUDA 版本冲突，又耗掉 4 天，进度条才到 30%。  \n- 训练出的模型风格强度不可控，FID 指标比论文高 15%，却找不到官方超参与训练日志，只能盲目调参。  \n- 答辩 PPT 里放不出可视化对比图，导师一句“复现可信度？”让小赵当场语塞。\n\n### 使用 Deep-Learning-Paper-Review-and-Practice 后\n- 打开仓库首页，直接锁定 ICCV 2017《Arbitrary Style Transfer in Real-time with Adaptive Instance Normalization》条目，阅读顺序一目了然。  \n- 30 分钟看完配套中文 PDF 总结 + 20 分钟 YouTube 视频，AdaIN 的均值方差对齐机制瞬间清晰，省下 2 天半。  \n- 一键 clone 代码实践 notebook，Colab 环境已配好，GPU 秒开；跑通 COCO + WikiArt 数据集，2 小时完成首轮训练。  \n- 仓库附带作者超参与训练日志，按表调 batch-size 与风格权重 λ，FID 从 45 降到 29，逼近论文 27.6。  \n- 用 notebook 里的可视化脚本生成 10 组风格迁移 GIF，直接插入 PPT，导师看完点头：“复现扎实，改进空间明确。”\n\nDeep-Learning-Paper-Review-and-Practice 让小赵用 5 天完成原本 3 周的工作量，把“读论文”变成“跑论文”，科研效率肉眼可见。","https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002Fndb796_Deep-Learning-Paper-Review-and-Practice_eb4a186a.png","ndb796","Dongbin Na \u002F 나동빈","https:\u002F\u002Foss.gittoolsai.com\u002Favatars\u002Fndb796_aea09709.jpg","Artificial Intelligence & Deep Learning\r\nComputer Security","POSTECH","Korea, Republic of","dongbinna@postech.ac.kr",null,"https:\u002F\u002Fgithub.com\u002Fndb796",[85],{"name":86,"color":87,"percentage":88},"Jupyter Notebook","#DA5B0B",100,1155,335,"2026-04-04T13:10:26","未说明",{"notes":94,"python":92,"dependencies":95},"README 中未给出任何运行环境或依赖信息，仅提供论文链接、视频讲解与若干 Jupyter Notebook 示例代码，需自行查看各 notebook 的 import 语句推断所需库版本",[92],[13,15,14],[98,99,100],"deep-learning","ai","paper-reviews","2026-03-27T02:49:30.150509","2026-04-06T06:46:16.160768",[104,109],{"id":105,"question_zh":106,"answer_zh":107,"source_url":108},6039,"代码中的损失函数为什么没有论文里的 Dissim 项？","这是 1-channel 攻击版本的实现。在 1-channel 攻击中，三个通道的值被强制设为相同，因此 Dissim 函数输出恒为 0，故省略。代码中通过 get_ct() 函数把单通道值复制到 R\u002FG\u002FB 三通道，使扰动更像自然阴影，但攻击强度会比 3-channel 版本弱。","https:\u002F\u002Fgithub.com\u002Fndb796\u002FDeep-Learning-Paper-Review-and-Practice\u002Fissues\u002F1",{"id":110,"question_zh":111,"answer_zh":112,"source_url":113},6040,"Style_Transfer_Tutorial 在 multiprocessing 时卡在 image_loader 的哪一步？","Issue 中未给出具体解决方案，但卡住的位置是 image = loader(image).unsqueeze(0)。常见原因包括：1) DataLoader 内部使用了多线程\u002F多进程导致死锁，可尝试把 num_workers 设为 0；2) 共享 CUDA 张量时未使用正确的进程启动方式（spawn\u002Fforkserver）。建议先用单进程调试，确认 loader 本身无误后再逐步引入 multiprocessing。","https:\u002F\u002Fgithub.com\u002Fndb796\u002FDeep-Learning-Paper-Review-and-Practice\u002Fissues\u002F4",[]]