{"id":30105212,"url":"https://github.com/liu42/inscriptions","last_synced_at":"2025-08-09T23:49:16.288Z","repository":{"id":253073107,"uuid":"842375543","full_name":"LIU42/Inscriptions","owner":"LIU42","description":"2024 年 MathorCup 数学应用挑战赛 B 题，基于 YOLOv8 的甲骨文原始拓片图像单字分割识别模型。","archived":false,"fork":false,"pushed_at":"2025-06-29T15:12:04.000Z","size":360,"stargazers_count":26,"open_issues_count":4,"forks_count":3,"subscribers_count":1,"default_branch":"main","last_synced_at":"2025-07-16T19:39:56.402Z","etag":null,"topics":["computer-vision","flask","image-classification","object-detection","onnxruntime","opencv","webapp","yolo","yolov8"],"latest_commit_sha":null,"homepage":"","language":"JavaScript","has_issues":true,"has_wiki":null,"has_pages":null,"mirror_url":null,"source_name":null,"license":null,"status":null,"scm":"git","pull_requests_enabled":true,"icon_url":"https://github.com/LIU42.png","metadata":{"files":{"readme":"README.md","changelog":null,"contributing":null,"funding":null,"license":null,"code_of_conduct":null,"threat_model":null,"audit":null,"citation":null,"codeowners":null,"security":null,"support":null,"governance":null,"roadmap":null,"authors":null,"dei":null,"publiccode":null,"codemeta":null,"zenodo":null}},"created_at":"2024-08-14T08:16:51.000Z","updated_at":"2025-06-29T15:04:52.000Z","dependencies_parsed_at":"2025-07-17T01:41:16.763Z","dependency_job_id":"399dc6e2-2b49-475b-8b18-ea82a74400e4","html_url":"https://github.com/LIU42/Inscriptions","commit_stats":null,"previous_names":["liu42/bonescripts","liu42/oraclebones","liu42/oracledetection","liu42/oraclerecognition","liu42/inscriptions"],"tags_count":3,"template":false,"template_full_name":null,"purl":"pkg:github/LIU42/Inscriptions","repository_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/LIU42%2FInscriptions","tags_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/LIU42%2FInscriptions/tags","releases_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/LIU42%2FInscriptions/releases","manifests_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/LIU42%2FInscriptions/manifests","owner_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/owners/LIU42","download_url":"https://codeload.github.com/LIU42/Inscriptions/tar.gz/refs/heads/main","sbom_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/LIU42%2FInscriptions/sbom","host":{"name":"GitHub","url":"https://github.com","kind":"github","repositories_count":269654601,"owners_count":24454326,"icon_url":"https://github.com/github.png","version":null,"created_at":"2022-05-30T11:31:42.601Z","updated_at":"2022-07-04T15:15:14.044Z","status":"online","status_checked_at":"2025-08-09T02:00:10.424Z","response_time":111,"last_error":null,"robots_txt_status":"success","robots_txt_updated_at":"2025-07-24T06:49:26.215Z","robots_txt_url":"https://github.com/robots.txt","online":true,"can_crawl_api":true,"host_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub","repositories_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories","repository_names_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repository_names","owners_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/owners"}},"keywords":["computer-vision","flask","image-classification","object-detection","onnxruntime","opencv","webapp","yolo","yolov8"],"created_at":"2025-08-09T23:49:15.677Z","updated_at":"2025-08-09T23:49:16.243Z","avatar_url":"https://github.com/LIU42.png","language":"JavaScript","funding_links":[],"categories":[],"sub_categories":[],"readme":"# Inscriptions\n\n*\u003cu\u003ev2.0.0 新变化：项目使用 C/S 架构重新实现。\u003c/u\u003e*\n\n## 项目简介\n\n本项目取材自 2024 年 MathorCup 数学应用挑战赛 B 题，基于 YOLOv8 的甲骨文原始拓片图像单字分割识别系统。针对原始拓片图像中的甲骨文文字分割识别包括以下两个阶段：\n\n- 目标检测：基于 YOLOv8 目标检测模型，对甲骨文文字所在的矩形区域进行提取。\n\n- 字符识别：基于 YOLOv8 图像分类模型，对文字图像进行分类，判断该字形所代表的具体字符内容。\n\n同时实现基于 [Flask](https://flask.palletsprojects.com/en/stable/) 的后端推理接口和 Web UI 可视化界面。\n\n## 效果展示\n\n![效果展示](./assets/example.jpg \"效果展示\")\n\n## 性能评估\n\n本项目训练数据集来自 [殷契文渊](https://jgw.aynu.edu.cn/)，对原始数据集进行了增强，采用 YOLOv8s 模型进行训练，能识别 其中195 中不同的甲骨文字符，其中目标检测模型各项指标如下：\n\n| mAP50 | mAP50-95 | Precision | Recall |\n|:-----:|:--------:|:---------:|:------:|\n| 0.928 | 0.608    | 0.905     | 0.865  |\n\n字符分类模型各项指标如下：\n\n| Top1_acc | Top5_acc |\n|:--------:|:--------:|\n| 0.897    | 0.969    |\n\n## 使用说明\n\n首先需要安装本项目依赖的各种库和工具包。\n\n```shell-session\npip install -r requirements.txt\n```\n\n可以在本项目 Releases 中下载我训练好的模型权重文件，解压到目录 inferences/models 下，运行以下命令以启动服务端程序。\n\n```shell-session\npython -m flask --app servers.server:app run --host=0.0.0.0 --port=8080\n```\n\n本项目识别程序默认的配置文件为 inferences/configs/config.toml，其中各个字段的描述如下。\n\n| 字段名                       | 字段描述                                      |\n|:-------------------------:|:-----------------------------------------:|\n| providers                 | 模型推理 ONNX Runtime Execution Providers 列表。 |\n| precision                 | 推理运算精度，可取 \"fp32\"（单精度）或 \"fp16\"（半精度）。       |\n| detection-model-path      | 目标检测模型加载路径。                               |\n| classification-model-path | 字符识别模型加载路径。                               |\n| conf-threshold            | 目标检测置信度阈值。                                |\n| iou-threshold             | 目标检测非极大值抑制 IoU 阈值。                        |\n\n客户端程序位于 clients 目录下，可通过 [Nginx](https://nginx.org/en/) 或其他前端服务器部署，其中 clients/js/names.js 内为字符类别名列表，可根据实际需要调整。此外还需要安装配置并启动 [Nginx](https://nginx.org/en/) 服务进行后端服务和前端服务之间的反向代理。\n\n如果需要使用自己的数据集训练模型，则需要安装 Ultralytics 框架，参照 [Ultralytics 官方文档](https://docs.ultralytics.com/) 进行模型的训练，最后将模型转换为 ONNX 格式进行部署即可。\n\n```shell-session\npip install ultralytics\n```\n\n\n","project_url":"https://awesome.ecosyste.ms/api/v1/projects/github.com%2Fliu42%2Finscriptions","html_url":"https://awesome.ecosyste.ms/projects/github.com%2Fliu42%2Finscriptions","lists_url":"https://awesome.ecosyste.ms/api/v1/projects/github.com%2Fliu42%2Finscriptions/lists"}