[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"similar-ZhugeKongan--torch-template-for-deep-learning":3,"tool-ZhugeKongan--torch-template-for-deep-learning":61},[4,18,26,36,44,53],{"id":5,"name":6,"github_repo":7,"description_zh":8,"stars":9,"difficulty_score":10,"last_commit_at":11,"category_tags":12,"status":17},4358,"openclaw","openclaw\u002Fopenclaw","OpenClaw 是一款专为个人打造的本地化 AI 助手，旨在让你在自己的设备上拥有完全可控的智能伙伴。它打破了传统 AI 助手局限于特定网页或应用的束缚，能够直接接入你日常使用的各类通讯渠道，包括微信、WhatsApp、Telegram、Discord、iMessage 等数十种平台。无论你在哪个聊天软件中发送消息，OpenClaw 都能即时响应，甚至支持在 macOS、iOS 和 Android 设备上进行语音交互，并提供实时的画布渲染功能供你操控。\n\n这款工具主要解决了用户对数据隐私、响应速度以及“始终在线”体验的需求。通过将 AI 部署在本地，用户无需依赖云端服务即可享受快速、私密的智能辅助，真正实现了“你的数据，你做主”。其独特的技术亮点在于强大的网关架构，将控制平面与核心助手分离，确保跨平台通信的流畅性与扩展性。\n\nOpenClaw 非常适合希望构建个性化工作流的技术爱好者、开发者，以及注重隐私保护且不愿被单一生态绑定的普通用户。只要具备基础的终端操作能力（支持 macOS、Linux 及 Windows WSL2），即可通过简单的命令行引导完成部署。如果你渴望拥有一个懂你",349277,3,"2026-04-06T06:32:30",[13,14,15,16],"Agent","开发框架","图像","数据工具","ready",{"id":19,"name":20,"github_repo":21,"description_zh":22,"stars":23,"difficulty_score":10,"last_commit_at":24,"category_tags":25,"status":17},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,"2026-04-05T11:01:52",[14,15,13],{"id":27,"name":28,"github_repo":29,"description_zh":30,"stars":31,"difficulty_score":32,"last_commit_at":33,"category_tags":34,"status":17},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",160015,2,"2026-04-18T11:30:52",[14,13,35],"语言模型",{"id":37,"name":38,"github_repo":39,"description_zh":40,"stars":41,"difficulty_score":32,"last_commit_at":42,"category_tags":43,"status":17},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。",109154,"2026-04-18T11:18:24",[14,15,13],{"id":45,"name":46,"github_repo":47,"description_zh":48,"stars":49,"difficulty_score":32,"last_commit_at":50,"category_tags":51,"status":17},6121,"gemini-cli","google-gemini\u002Fgemini-cli","gemini-cli 是一款由谷歌推出的开源 AI 命令行工具，它将强大的 Gemini 大模型能力直接集成到用户的终端环境中。对于习惯在命令行工作的开发者而言，它提供了一条从输入提示词到获取模型响应的最短路径，无需切换窗口即可享受智能辅助。\n\n这款工具主要解决了开发过程中频繁上下文切换的痛点，让用户能在熟悉的终端界面内直接完成代码理解、生成、调试以及自动化运维任务。无论是查询大型代码库、根据草图生成应用，还是执行复杂的 Git 操作，gemini-cli 都能通过自然语言指令高效处理。\n\n它特别适合广大软件工程师、DevOps 人员及技术研究人员使用。其核心亮点包括支持高达 100 万 token 的超长上下文窗口，具备出色的逻辑推理能力；内置 Google 搜索、文件操作及 Shell 命令执行等实用工具；更独特的是，它支持 MCP（模型上下文协议），允许用户灵活扩展自定义集成，连接如图像生成等外部能力。此外，个人谷歌账号即可享受免费的额度支持，且项目基于 Apache 2.0 协议完全开源，是提升终端工作效率的理想助手。",100752,"2026-04-10T01:20:03",[52,13,15,14],"插件",{"id":54,"name":55,"github_repo":56,"description_zh":57,"stars":58,"difficulty_score":32,"last_commit_at":59,"category_tags":60,"status":17},4721,"markitdown","microsoft\u002Fmarkitdown","MarkItDown 是一款由微软 AutoGen 团队打造的轻量级 Python 工具，专为将各类文件高效转换为 Markdown 格式而设计。它支持 PDF、Word、Excel、PPT、图片（含 OCR）、音频（含语音转录）、HTML 乃至 YouTube 链接等多种格式的解析，能够精准提取文档中的标题、列表、表格和链接等关键结构信息。\n\n在人工智能应用日益普及的今天，大语言模型（LLM）虽擅长处理文本，却难以直接读取复杂的二进制办公文档。MarkItDown 恰好解决了这一痛点，它将非结构化或半结构化的文件转化为模型“原生理解”且 Token 效率极高的 Markdown 格式，成为连接本地文件与 AI 分析 pipeline 的理想桥梁。此外，它还提供了 MCP（模型上下文协议）服务器，可无缝集成到 Claude Desktop 等 LLM 应用中。\n\n这款工具特别适合开发者、数据科学家及 AI 研究人员使用，尤其是那些需要构建文档检索增强生成（RAG）系统、进行批量文本分析或希望让 AI 助手直接“阅读”本地文件的用户。虽然生成的内容也具备一定可读性，但其核心优势在于为机器",93400,"2026-04-06T19:52:38",[52,14],{"id":62,"github_repo":63,"name":64,"description_en":65,"description_zh":66,"ai_summary_zh":66,"readme_en":67,"readme_zh":68,"quickstart_zh":69,"use_case_zh":70,"hero_image_url":71,"owner_login":72,"owner_name":73,"owner_avatar_url":74,"owner_bio":75,"owner_company":76,"owner_location":76,"owner_email":76,"owner_twitter":76,"owner_website":76,"owner_url":77,"languages":78,"stars":98,"forks":99,"last_commit_at":100,"license":101,"difficulty_score":32,"env_os":102,"env_gpu":103,"env_ram":103,"env_deps":104,"category_tags":110,"github_topics":76,"view_count":32,"oss_zip_url":76,"oss_zip_packed_at":76,"status":17,"created_at":111,"updated_at":112,"faqs":113,"releases":114},9310,"ZhugeKongan\u002Ftorch-template-for-deep-learning","torch-template-for-deep-learning","Pytorch Implementations of large number  classical backbone CNNs, data enhancement, torch loss, attention, visualization and  some common algorithms.","torch-template-for-deep-learning 是一个基于 PyTorch 的深度学习综合模板库，旨在为开发者提供一站式的算法实现与训练框架。它解决了研究人员和工程师在复现经典模型、尝试新型注意力机制或配置复杂数据增强策略时，需要重复编写基础代码的痛点，显著提升了实验效率。\n\n该工具非常适合深度学习研究者、算法工程师以及希望快速上手 PyTorch 项目的学生使用。其核心亮点在于极高的集成度：不仅涵盖了从 AlexNet、ResNet 到 EfficientNet 等数十种经典骨干网络，还集中实现了包括 SE、CBAM、CoAtNet 在内的三十多种前沿注意力机制。此外，它内置了 Mixup、Cutmix、StochDepth 等多种高级数据增强与正则化技术，并提供了清晰的数据加载器示例。\n\n除了算法丰富，torch-template-for-deep-learning 还贴心地准备了完整的训练脚本基线，并支持 Web 端与 C++ 两种模型部署模式，帮助用户轻松完成从模型训练到落地应用的全流程。无论是用于学术探索还是工程原型开发，它都是一个实用且高效的起点。","# Torch-template-for-deep-learning\n Pytorch implementations of some **classical backbone CNNs, data enhancement, torch loss, attention, visualization and  some common algorithms **.\n\n### Requirements\n\n  · torch, torch-vision\n\n  · torchsummary\n  \n  · other necessary\n\n### usage\nA training script is supplied in “train_baseline.py”, the arguments are in “args.py\n\u003Cimg src=\"https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002FZhugeKongan_torch-template-for-deep-learning_readme_ebc5d81c2b7c.png\" width=80% \u002F>\n### autoaug: Data enhancement and CNNs regularization\n    - StochDepth\n    - label smoothing\n    - Cutout\n    - DropBlock\n    - Mixup\n    - Manifold Mixup\n    - ShakeDrop\n    - cutmix\n### dataset_loader: Loaders for various datasets\n```python\nfrom dataloder.scoliosis_dataloder import ScoliosisDataset\nfrom dataloder.facial_attraction_dataloder import FacialAttractionDataset\nfrom dataloder.fa_and_sco_dataloder import ScoandFaDataset\nfrom dataloder.scofaNshot_dataloder import ScoandFaNshotDataset\nfrom dataloder.age_dataloder import MegaAsiaAgeDataset\ndef load_dataset(data_config):\n    if data_config.dataset == 'cifar10':\n        training_transform=training_transforms()\n        if data_config.autoaug:\n            print('auto Augmentation the data !')\n            training_transform.transforms.insert(0, Augmentation(fa_reduced_cifar10()))\n        train_dataset = torchvision.datasets.CIFAR10(root=data_config.data_path,\n                                                     train=True,\n                                                     transform=training_transform,\n                                                     download=True)\n        val_dataset = torchvision.datasets.CIFAR10(root=data_config.data_path,\n                                                   train=False,\n                                                   transform=validation_transforms(),\n                                                   download=True)\n        return train_dataset,val_dataset\n    elif data_config.dataset == 'cifar100':\n        train_dataset = torchvision.datasets.CIFAR100(root=data_config.data_path,\n                                                     train=True,\n                                                     transform=training_transforms(),\n                                                     download=True)\n        val_dataset = torchvision.datasets.CIFAR100(root=data_config.data_path,\n                                                   train=False,\n                                                   transform=validation_transforms(),\n                                                   download=True)\n        return train_dataset, val_dataset\n```\n### deployment: Deployment mode of pytorch model\n    Two deployment modes of pytorch model are provided, one is web deployment and the other is C + + deployment\n\n    Store the training weight file in ` flash_ Deployment ` folder\n\n    Then modify ' server.py '  path\n\n    Leverage ' client.Py ' call\n### models: Various classical deep learning models\n\n##### Classical network\n    - **AlexNet**\n    - **VGG**\n    - **ResNet** \n    - **ResNext** \n    - **InceptionV1**\n    - **InceptionV2 and InceptionV3**\n    - **InceptionV4 and Inception-ResNet**\n    - **GoogleNet**\n    - **EfficienNet**\n    - **MNasNet**\n    - **DPN**\n\n##### Attention network\n    - **SE Attention**\n    - **External Attention**\n    - **Self Attention**\n    - **SK Attention**\n    - **CBAM Attention**\n    - **BAM Attention**\n    - **ECA Attention**\n    - **DANet Attention**\n    - **Pyramid Split Attention(PSA)**\n    - **EMSA Attention**\n    - **A2Attention**\n    - **Non-Local Attention**\n    - **CoAtNet**\n    - **CoordAttention**\n    - **HaloAttention**\n    - **MobileViTAttention**\n    - **MUSEAttention**  \n    - **OutlookAttention**\n    - **ParNetAttention**\n    - **ParallelPolarizedSelfAttention**\n    - **residual_attention**\n    - **S2Attention**\n    - **SpatialGroupEnhance Attention**\n    - **ShuffleAttention**\n    - **GFNet Attention**\n    - **TripletAttention**\n    - **UFOAttention**\n    - **VIPAttention**\n\n##### Lightweight network\n    - **MobileNets:**\n    - **MobileNetV2：**\n    - **MobileNetV3：**\n    - **ShuffleNet：**\n    - **ShuffleNet V2:**\n    - **SqueezeNet**\n    - **Xception**\n    - **MixNet**\n    - **GhostNet**\n    \n##### GAN\n    - **Auxiliary Classifier GAN**\n    - **Adversarial Autoencoder**\n    - **BEGAN**\n    - **BicycleGAN**\n    - **Boundary-Seeking GAN**\n    - **Cluster GAN**\n    - **Conditional GAN**\n    - **Context-Conditional GAN**\n    - **Context Encoder**\n    - **Coupled GAN**\n    - **CycleGAN**\n    - **Deep Convolutional GAN**\n    - **DiscoGAN**\n    - **DRAGAN**\n    - **DualGAN**\n    - **Energy-Based GAN**\n    - **Enhanced Super-Resolution GAN**  \n    - **Least Squares GAN**\n    - **Enhanced Super-Resolution GAN**\n    - **GAN**\n    - **InfoGAN**\n    - **Pix2Pix**\n    - **Relativistic GAN**\n    - **Semi-Supervised GAN**\n    - **StarGAN**\n    - **Wasserstein GAN**\n    - **Wasserstein GAN GP**\n    - **Wasserstein GAN DIV**\n\n##### ObjectDetection-network\n\n    - **SSD:**\n    - **YOLO:**\n    - **YOLOv2:**\n    - **YOLOv3:**\n    - **FCOS:**\n    - **FPN:**\n    - **RetinaNet**\n    - **Objects as Points:**\n    - **FSAF:**\n    - **CenterNet**\n    - **FoveaBox**\n\n##### Semantic Segmentation\n\n    - **FCN**\n    - **Fast-SCNN**\n    - **LEDNet:**\n    - **LRNNet**\n    - **FisheyeMODNet:**\n  \n##### Instance Segmentation\n    - **PolarMask** \n  \n##### FaceDetectorAndRecognition\n    - **FaceBoxes**\n    - **LFFD**\n    - **VarGFaceNet**\n\n##### HumanPoseEstimation\n\n    - **Stacked Hourglass Networks**\n    - **Simple Baselines**\n    - **LPN**\n    \n### pytorch_loss: loss for training\n    - label-smooth\n    - amsoftmax\n    - focal-loss\n    - dual-focal-loss \n    - triplet-loss\n    - giou-loss\n    - affinity-loss\n    - pc_softmax_cross_entropy\n    - ohem-loss(softmax based on line hard mining loss)\n    - large-margin-softmax(bmvc2019)\n    - lovasz-softmax-loss\n    - dice-loss(both generalized soft dice loss and batch soft dice loss)\n\n### tf_to_pytorch: TensorFlow to PyTorch Conversion\n    This directory is used to convert TensorFlow weights to PyTorch. \n    It was hacked together fairly quickly, so the code is not the most \n    beautiful (just a warning!), but it does the job. I will be refactoring it soon.\n\n### TorchCAM: Class Activation Mapping\n    Simple way to leverage the class-specific activation of convolutional layers in PyTorch.\n    \n    - CAM\n    - ScoreCAM\n    - SSCAM\n    - ISCAM\n    - GradCAM\n    - Grad-CAM++\n    - Smooth Grad-CAM++\n    - XGradCAM\n    - LayerCAM\n    \n    \n### Note\n- **More modules may be added later**.\n\n- **During the implementation process, I read a lot of codes and articles and referred to a lot of contents.\n     Some have added copyright notices, and some don't remember the main references. If there is infringement, please contact to delete.**\n\n- **I wrote some blogs（which are in Chinese） to introduce the models implemented in this project**：\n    - [torch模板使用说明](https:\u002F\u002Fmp.weixin.qq.com\u002Fs\u002FnXiFM6Mila2bH7fopbsVOw)\n    - [论文综述：注意力机制](https:\u002F\u002Fzhuanlan.zhihu.com\u002Fp\u002F388122250)\n    - [论文综述：特征可视化](https:\u002F\u002Fzhuanlan.zhihu.com\u002Fp\u002F420954745)\n    - [论文综述：数据增强&网络正则化](https:\u002F\u002Fzhuanlan.zhihu.com\u002Fp\u002F402511359)\n    - [论文综述：轻量型网络](https:\u002F\u002Fmp.weixin.qq.com\u002Fs\u002Fw9XKRzkxNfmNjUdlVuEyTQ)\n    \n    \n- **Some of My Reference Repositories**：\n    - https:\u002F\u002Fgithub.com\u002Fxmu-xiaoma666\u002FExternal-Attention-pytorch\n    - https:\u002F\u002Fgithub.com\u002Feriklindernoren\u002FPyTorch-GAN\n    - https:\u002F\u002Fwww.zhihu.com\u002Fpeople\u002FZhugeKongan\n    - https:\u002F\u002Fgithub.com\u002FZhugeKongan\u002FAttention-mechanism-implementation\n    - https:\u002F\u002Fgithub.com\u002FZhugeKongan\u002F-DataAug-and-NetRegularization\n    \n\n## Write at the end\nAt present, the work organized by this project is indeed not comprehensive enough. As the amount of reading increases, we will continue to improve this project. Welcome everyone star to support. If there are incorrect statements or incorrect code implementations in the article, you are welcome to point out~\n\n \n","# 深度学习PyTorch模板\n一些**经典骨干CNN、数据增强、PyTorch损失函数、注意力机制、可视化及常用算法**的PyTorch实现。\n\n### 需求\n  · torch, torch-vision\n\n  · torchsummary\n  \n  · 其他必要依赖\n\n### 使用方法\n在“train_baseline.py”中提供了一个训练脚本，参数定义在“args.py”中。\n\u003Cimg src=\"https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002FZhugeKongan_torch-template-for-deep-learning_readme_ebc5d81c2b7c.png\" width=80% \u002F>\n### autoaug: 数据增强与CNN正则化\n    - StochDepth\n    - 标签平滑\n    - Cutout\n    - DropBlock\n    - Mixup\n    - Manifold Mixup\n    - ShakeDrop\n    - cutmix\n### dataset_loader: 各类数据集加载器\n```python\nfrom dataloder.scoliosis_dataloder import ScoliosisDataset\nfrom dataloder.facial_attraction_dataloder import FacialAttractionDataset\nfrom dataloder.fa_and_sco_dataloder import ScoandFaDataset\nfrom dataloder.scofaNshot_dataloder import ScoandFaNshotDataset\nfrom dataloder.age_dataloder import MegaAsiaAgeDataset\ndef load_dataset(data_config):\n    if data_config.dataset == 'cifar10':\n        training_transform=training_transforms()\n        if data_config.autoaug:\n            print('自动增强数据！')\n            training_transform.transforms.insert(0, Augmentation(fa_reduced_cifar10()))\n        train_dataset = torchvision.datasets.CIFAR10(root=data_config.data_path,\n                                                     train=True,\n                                                     transform=training_transform,\n                                                     download=True)\n        val_dataset = torchvision.datasets.CIFAR10(root=data_config.data_path,\n                                                   train=False,\n                                                   transform=validation_transforms(),\n                                                   download=True)\n        return train_dataset,val_dataset\n    elif data_config.dataset == 'cifar100':\n        train_dataset = torchvision.datasets.CIFAR100(root=data_config.data_path,\n                                                     train=True,\n                                                     transform=training_transforms(),\n                                                     download=True)\n        val_dataset = torchvision.datasets.CIFAR100(root=data_config.data_path,\n                                                   train=False,\n                                                   transform=validation_transforms(),\n                                                   download=True)\n        return train_dataset, val_dataset\n```\n### deployment: PyTorch模型部署模式\n    提供了两种PyTorch模型的部署方式，一种是Web部署，另一种是C++部署。\n\n    将训练好的权重文件存放在`flash_ Deployment`文件夹中。\n\n    然后修改‘server.py’中的路径。\n\n    调用‘client.Py’进行使用。\n### models: 各种经典深度学习模型\n\n##### 经典网络\n    - **AlexNet**\n    - **VGG**\n    - **ResNet** \n    - **ResNext** \n    - **InceptionV1**\n    - **InceptionV2和InceptionV3**\n    - **InceptionV4和Inception-ResNet**\n    - **GoogleNet**\n    - **EfficienNet**\n    - **MNasNet**\n    - **DPN**\n\n##### 注意力网络\n    - **SE Attention**\n    - **External Attention**\n    - **Self Attention**\n    - **SK Attention**\n    - **CBAM Attention**\n    - **BAM Attention**\n    - **ECA Attention**\n    - **DANet Attention**\n    - **Pyramid Split Attention(PSA)**\n    - **EMSA Attention**\n    - **A2Attention**\n    - **Non-Local Attention**\n    - **CoAtNet**\n    - **CoordAttention**\n    - **HaloAttention**\n    - **MobileViTAttention**\n    - **MUSEAttention**  \n    - **OutlookAttention**\n    - **ParNetAttention**\n    - **ParallelPolarizedSelfAttention**\n    - **residual_attention**\n    - **S2Attention**\n    - **SpatialGroupEnhance Attention**\n    - **ShuffleAttention**\n    - **GFNet Attention**\n    - **TripletAttention**\n    - **UFOAttention**\n    - **VIPAttention**\n\n##### 轻量级网络\n    - **MobileNets:**\n    - **MobileNetV2：**\n    - **MobileNetV3：**\n    - **ShuffleNet：**\n    - **ShuffleNet V2:**\n    - **SqueezeNet**\n    - **Xception**\n    - **MixNet**\n    - **GhostNet**\n    \n##### GAN\n    - **辅助分类器GAN**\n    - **对抗自编码器**\n    - **BEGAN**\n    - **BicycleGAN**\n    - **边界搜索GAN**\n    - **聚类GAN**\n    - **条件GAN**\n    - **上下文条件GAN**\n    - **上下文编码器**\n    - **耦合GAN**\n    - **CycleGAN**\n    - **深度卷积GAN**\n    - **DiscoGAN**\n    - **DRAGAN**\n    - **DualGAN**\n    - **基于能量的GAN**\n    - **增强型超分辨率GAN**  \n    - **最小二乘GAN**\n    - **InfoGAN**\n    - **Pix2Pix**\n    - **相对论GAN**\n    - **半监督GAN**\n    - **StarGAN**\n    - **Wasserstein GAN**\n    - **Wasserstein GAN GP**\n    - **Wasserstein GAN DIV**\n\n##### 目标检测网络\n\n    - **SSD:**\n    - **YOLO:**\n    - **YOLOv2:**\n    - **YOLOv3:**\n    - **FCOS:**\n    - **FPN:**\n    - **RetinaNet**\n    - **以点表示目标:**\n    - **FSAF:**\n    - **CenterNet**\n    - **FoveaBox**\n\n##### 语义分割\n\n    - **FCN**\n    - **Fast-SCNN**\n    - **LEDNet:**\n    - **LRNNet**\n    - **FisheyeMODNet:**\n\n##### 实例分割\n    - **PolarMask** \n  \n##### 人脸检测与识别\n    - **FaceBoxes**\n    - **LFFD**\n    - **VarGFaceNet**\n\n##### 人体姿态估计\n\n    - **堆叠沙漏网络**\n    - **简单基线**\n    - **LPN**\n    \n### pytorch_loss: 训练用损失函数\n    - 标签平滑\n    - amsoftmax\n    - focal-loss\n    - 双重focal-loss \n    - triplet-loss\n    - giou-loss\n    - affinity-loss\n    - pc_softmax_cross_entropy\n    - ohem-loss(基于线性硬挖掘的softmax损失)\n    - 大间隔softmax(bmvc2019)\n    - lovasz-softmax-loss\n    - dice-loss(包括广义软骰子损失和批次软骰子损失)\n\n### tf_to_pytorch: TensorFlow到PyTorch转换\n    该目录用于将TensorFlow权重转换为PyTorch格式。由于是快速拼凑而成，代码并不十分优雅（仅供参考），但功能上可以满足需求。后续会对其进行重构。\n\n### TorchCAM: 类激活映射\n    在PyTorch中简单地利用卷积层的类别特定激活的一种方法。\n    \n    - CAM\n    - ScoreCAM\n    - SSCAM\n    - ISCAM\n    - GradCAM\n    - Grad-CAM++\n    - Smooth Grad-CAM++\n    - XGradCAM\n    - LayerCAM\n\n### 注\n- **后续可能会添加更多模块**。\n\n- **在实现过程中，我阅读了大量的代码和文章，并参考了许多资料。\n     其中一些已添加版权声明，而另一些则记不清主要参考来源。如有侵权，请联系删除。**\n\n- **我撰写了一些博客（均为中文），用于介绍本项目中实现的模型**：\n    - [torch模板使用说明](https:\u002F\u002Fmp.weixin.qq.com\u002Fs\u002FnXiFM6Mila2bH7fopbsVOw)\n    - [论文综述：注意力机制](https:\u002F\u002Fzhuanlan.zhihu.com\u002Fp\u002F388122250)\n    - [论文综述：特征可视化](https:\u002F\u002Fzhuanlan.zhihu.com\u002Fp\u002F420954745)\n    - [论文综述：数据增强&网络正则化](https:\u002F\u002Fzhuanlan.zhihu.com\u002Fp\u002F402511359)\n    - [论文综述：轻量型网络](https:\u002F\u002Fmp.weixin.qq.com\u002Fs\u002Fw9XKRzkxNfmNjUdlVuEyTQ)\n    \n    \n- **我的部分参考仓库**：\n    - https:\u002F\u002Fgithub.com\u002Fxmu-xiaoma666\u002FExternal-Attention-pytorch\n    - https:\u002F\u002Fgithub.com\u002Feriklindernoren\u002FPyTorch-GAN\n    - https:\u002F\u002Fwww.zhihu.com\u002Fpeople\u002FZhugeKongan\n    - https:\u002F\u002Fgithub.com\u002FZhugeKongan\u002FAttention-mechanism-implementation\n    - https:\u002F\u002Fgithub.com\u002FZhugeKongan\u002F-DataAug-and-NetRegularization\n    \n\n## 结尾语\n目前，本项目整理的内容确实还不够全面。随着阅读量的增加，我们将继续完善该项目。欢迎大家点赞支持。如果文中存在表述错误或代码实现不当之处，也欢迎指出~","# torch-template-for-deep-learning 快速上手指南\n\n本工具是一个基于 PyTorch 的深度学习模板库，集成了经典骨干网络、数据增强策略、多种损失函数、注意力机制、可视化方法及常见算法（如 GAN、目标检测、分割等），旨在帮助开发者快速搭建和实验深度学习模型。\n\n## 环境准备\n\n在开始之前，请确保您的系统满足以下要求：\n\n*   **操作系统**：Linux \u002F Windows \u002F macOS\n*   **Python 版本**：建议 Python 3.7+\n*   **核心依赖**：\n    *   `torch` (PyTorch)\n    *   `torchvision`\n    *   `torchsummary`\n    *   其他常规科学计算库 (如 `numpy`, `pillow` 等)\n\n> **国内加速建议**：推荐使用清华或阿里镜像源安装依赖，以提升下载速度。\n\n## 安装步骤\n\n1.  **克隆项目代码**\n    ```bash\n    git clone https:\u002F\u002Fgithub.com\u002FZhugeKongan\u002Ftorch-template-for-deep-learning.git\n    cd torch-template-for-deep-learning\n    ```\n\n2.  **安装依赖包**\n    \n    使用 pip 安装基础依赖（推荐配置国内镜像源）：\n    ```bash\n    pip install torch torchvision torchsummary -i https:\u002F\u002Fpypi.tuna.tsinghua.edu.cn\u002Fsimple\n    ```\n    \n    如果项目根目录下有 `requirements.txt` 文件，也可一键安装所有必要包：\n    ```bash\n    pip install -r requirements.txt -i https:\u002F\u002Fpypi.tuna.tsinghua.edu.cn\u002Fsimple\n    ```\n\n## 基本使用\n\n本项目提供了标准化的训练脚本和参数配置模块，最简单的使用流程如下：\n\n### 1. 配置训练参数\n所有的训练参数（如数据集路径、模型选择、超参数等）均在 `args.py` 文件中定义。您可以根据需求直接修改该文件，或通过命令行传递参数。\n\n### 2. 启动训练\n使用提供的 `train_baseline.py` 脚本开始训练。以下是一个典型的运行示例（假设使用 CIFAR-10 数据集和 ResNet 模型）：\n\n```bash\npython train_baseline.py --dataset cifar10 --model resnet18 --data_path .\u002Fdata\n```\n\n*   `--dataset`: 指定数据集名称（支持 `cifar10`, `cifar100` 及自定义数据集）。\n*   `--model`: 指定骨干网络（支持 AlexNet, VGG, ResNet, MobileNet, EfficientNet 等）。\n*   `--autoaug`: 若需启用自动数据增强（如 Cutout, Mixup, Label Smoothing 等），可添加此标志。\n\n### 3. 代码集成示例\n如果您希望在自定义代码中加载数据集并使用增强策略，可以参考以下模式：\n\n```python\nfrom dataloder.scoliosis_dataloder import ScoliosisDataset\n# 引入其他所需的数据集加载器\n\ndef load_dataset(data_config):\n    if data_config.dataset == 'cifar10':\n        training_transform = training_transforms()\n        # 启用自动数据增强\n        if data_config.autoaug:\n            print('auto Augmentation the data !')\n            training_transform.transforms.insert(0, Augmentation(fa_reduced_cifar10()))\n        \n        train_dataset = torchvision.datasets.CIFAR10(root=data_config.data_path,\n                                                     train=True,\n                                                     transform=training_transform,\n                                                     download=True)\n        val_dataset = torchvision.datasets.CIFAR10(root=data_config.data_path,\n                                                   train=False,\n                                                   transform=validation_transforms(),\n                                                   download=True)\n        return train_dataset, val_dataset\n    # 其他数据集逻辑...\n```\n\n### 4. 模型部署\n训练完成后，将权重文件存入 `flash_Deployment` 文件夹。\n*   **Web 部署**：修改 `server.py` 中的权重路径，运行服务器。\n*   **C++ 部署**：利用提供的接口进行调用。\n\n---\n*注：本项目包含丰富的预实现模型（如各类 Attention 机制、GAN 变体、YOLO 系列等），具体模型名称可在 `models` 目录中查阅并直接在 `args.py` 中调用。*","某医疗 AI 团队正在开发基于 X 光片的脊柱侧弯自动筛查系统，急需快速验证多种经典卷积神经网络与注意力机制的效果。\n\n### 没有 torch-template-for-deep-learning 时\n- **重复造轮子耗时**：工程师需手动从论文复现 ResNet、EfficientNet 等骨干网络及 SE、CBAM 等注意力模块，代码调试占据大量研发时间。\n- **数据增强策略单一**：缺乏 Mixup、Cutmix、Manifold Mixup 等高级正则化手段的集成实现，导致模型在少量医疗数据上容易过拟合，泛化能力差。\n- **部署流程割裂**：训练好的模型转为 Web 服务或 C++ 嵌入式部署时，需重新编写推理接口和预处理逻辑，环境配置复杂且易出错。\n- **实验管理混乱**：不同算法的实验脚本结构不统一，超参数调整和数据加载器（如特定的脊柱数据集 loader）需要反复修改，难以横向对比性能。\n\n### 使用 torch-template-for-deep-learning 后\n- **即插即用模型库**：直接调用内置的 AlexNet 到 CoAtNet 等数十种经典网络及注意力机制，将模型搭建时间从数天缩短至几分钟。\n- **一键启用高级增强**：通过配置文件轻松开启 AutoAugment、StochDepth 或 Label Smoothing，显著提升了小样本医疗影像的分类准确率。\n- **标准化部署方案**：利用提供的 Web 和 C++ 双模式部署模板，仅需替换权重文件并修改路径，即可快速完成从训练到临床原型的落地。\n- **统一实验框架**：复用标准的 `train_baseline.py` 和专用数据加载器（如 `ScoliosisDataset`），确保所有对比实验在相同基准下高效运行。\n\ntorch-template-for-deep-learning 通过提供一站式的算法实现与工程模板，让研发团队从繁琐的基础代码构建中解放出来，专注于核心业务逻辑的优化与迭代。","https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002FZhugeKongan_torch-template-for-deep-learning_d8011329.png","ZhugeKongan","Li Shengyan","https:\u002F\u002Foss.gittoolsai.com\u002Favatars\u002FZhugeKongan_3bdba851.jpg","960098070@qq.com\r\n",null,"https:\u002F\u002Fgithub.com\u002FZhugeKongan",[79,83,87,91,94],{"name":80,"color":81,"percentage":82},"Python","#3572A5",99.6,{"name":84,"color":85,"percentage":86},"Shell","#89e051",0.2,{"name":88,"color":89,"percentage":90},"C++","#f34b7d",0.1,{"name":92,"color":93,"percentage":90},"Lua","#000080",{"name":95,"color":96,"percentage":97},"CMake","#DA3434",0,1263,201,"2026-04-17T03:50:07","Apache-2.0","","未说明",{"notes":105,"python":103,"dependencies":106},"README 中仅列出了核心依赖库（torch, torchvision, torchsummary），未明确指定操作系统、GPU 型号、显存大小、内存需求及 Python 版本。该项目包含多种深度学习模型（如 GAN、目标检测等），实际运行时的硬件资源需求将取决于具体使用的模型和数据集规模。部署部分支持 Web 和 C++ 模式。",[107,108,109],"torch","torchvision","torchsummary",[15,14],"2026-03-27T02:49:30.150509","2026-04-19T06:02:10.371790",[],[]]