Ecosyste.ms: Awesome
An open API service indexing awesome lists of open source software.
https://github.com/evilgix/evil
Optical Character Recognition in Swift for iOS&macOS. 银行卡、身份证、门牌号光学识别
https://github.com/evilgix/evil
cnn-model keras machine-learning ocr swift4 vision
Last synced: 6 days ago
JSON representation
Optical Character Recognition in Swift for iOS&macOS. 银行卡、身份证、门牌号光学识别
- Host: GitHub
- URL: https://github.com/evilgix/evil
- Owner: evilgix
- License: mit
- Created: 2018-01-21T11:45:42.000Z (almost 7 years ago)
- Default Branch: master
- Last Pushed: 2018-02-14T05:43:56.000Z (over 6 years ago)
- Last Synced: 2024-10-13T14:41:08.317Z (23 days ago)
- Topics: cnn-model, keras, machine-learning, ocr, swift4, vision
- Language: Swift
- Homepage:
- Size: 3.13 MB
- Stars: 690
- Watchers: 21
- Forks: 99
- Open Issues: 3
-
Metadata Files:
- Readme: README.md
- License: LICENSE
Awesome Lists containing this project
README
Evil
Optical Character Recognition in Swift for iOS&macOS.
---
图片来自于网络，侵删
## Requirements
- iOS 11.0+ / macOS 10.13+
- Xcode9.2+
- Swift 4.0+
## [Installation](installation.md)
## Basic Usage
Evil 使用 [Vision](https://developer.apple.com/documentation/vision) 框架进行文字识别，因此你需要导入特定的[MLModel](https://developer.apple.com/documentation/coreml/mlmodel)到你的Xcode工程。例如识别身份证导入识别身份证的model，识别银行卡导入识别银行卡的model。
以下是目前已经训练好的模型
- ChineseIDCard ==> [🇨🇳中国二代身份证号码识别](http://ou5pk1mdu.bkt.clouddn.com/ChineseIDCard.mlmodel)
### 如何导入
首先从上面的地址下载或者自己训练得到模型文件(`*.mlmodel`)，然后作为resource资源加入项目。但是模型文件size一般都比较大因此Evil还支持另外一种方式即：运行时从指定的服务器下载对应的模型。
1. 获取对应的模型文件并重名命名为`[name].mlmodel`
`[name]`可通过以下方法获取`print(Recognizer.chineseIDCard.name)`
2. 将模型文件上传至自己的服务器或者CDN并获取下载链接
3. 在`info.plist` 中配置下载链接如下
### 初始化
``` Swift
lazy var evil = try? Evil(recognizer: .chineseIDCard)
```
或者 手动指定下载路径
``` Swift
let downloadURL = URL(string: "https://****/ChineseIDCard.mlmodel")!
let chineseIDCard = Recognizer.custom(name: "ChineseIDCard", model: downloadURL, needComplie: true, processor: Recognizer.cropChineseIDCardNumberArea)
let evil = try? Evil(recognizer: chineseIDCard)
```
或者直接将模型文件拖入 xcode 项目
``` Swift
let url = Bundle.main.url(forResource: "ChineseIDCard", withExtension: "mlmodelc")!
let evil = try? Evil(contentsOf: url, name: "ChineseIDCard")
```
注意Extension 为`mlmodelc`而不是`mlmodel`因为xcode会自动编译`mlmodel`文件，并且生成对应的`class`. 所以请务必注意Model文件不要和现有的`class`重名。
### 开始识别
``` Swift
let image: Recognizable = .....
let result = evil?.recognize(image)
print(result)
```
## 训练模型
参考Tools 目录下`PrepareBot`项目
参考博文[掘金](https://juejin.im/post/5a76ada6f265da4e9d220727)
## 交流讨论
1.github issue (推荐)
2.qq 群：641256202
## LICENSE
MIT
Copyright 2018 - 2019 Kevin.Gong aoxianglele#icloud.com