谷歌开源 Magika：毫秒级识别内容类型，百万文件测试准确率超 99%
近日，谷歌更新博文，宣布开源 Magika，这是一个基于人工智能的文件格式和内容类型识别工具。Magika 采用了一个定制的、高度优化的深度学习模型，即使在 CPU 上运行，也能在几毫秒内精确识别文件类型。
据悉，Magika 经过了百万文件级别的测试，准确率超过 99%。它支持识别超过 1500 种文件格式，包括图像、视频、音频、文档、压缩文件和可执行文件等。
谷歌表示，Magika 可以广泛应用于各种场景，例如文件管理、数据分类、安全分析和恶意软件检测等。它可以帮助用户快速识别和组织大量文件，提高工作效率。
值得一提的是，Magika 的开源意味着开发者可以自由使用和修改其代码，从而进一步扩展其功能和应用场景。
谷歌方面表示，Magika 的开源旨在促进人工智能在文件处理领域的创新和发展。它希望 Magika 能够成为一个社区驱动的项目，吸引更多的开发者参与其中，共同打造一个更加强大的文件识别工具。
目前，Magika 的相关源代码已托管到 GitHub 上，开发者可以免费获取和使用。
英语如下：
**Headline: Google Magika Debuts: Millisecond Content Recognition with 99%+ Accuracy**
**Keywords: AI Recognition, Open-Source Tool, File Classification**
**News Content:**
Google has open-sourced Magika, an AI-powered tool for file format and content type recognition. Magika employs a custom, highly optimized deep learning model that enables it to accurately identify file types within milliseconds, even when running on a CPU.
Magika has been tested on millions of files, achieving an accuracy rate of over 99%.It supports the recognition of over 1,500 file formats, including images, videos, audio, documents, compressed files, and executables.
Google notes that Magika has a wide range of potential applications, including file management, data classification, security analysis, and malware detection. It can help users quickly identify and organize large volumes of files, improving productivity.
Importantly, Magika’s open-source nature means that developers are free to use and modify its code, further extending its functionality and use cases.
Google says that Magika’s open-source release aims to foster innovation and advancement in the fieldof AI-powered file handling. It hopes that Magika will become a community-driven project, attracting more developers to contribute and build an even more robust file recognition tool.
Magika’s source code is now hosted on GitHub, where developers can access and use it for free.
【来源】https://www.ithome.com/0/750/474.htm
Views: 1