/Florence_2_tagger

an embedded package for Florence-2 for quick interrogator

MIT LicenseMIT

Florence_2_tagger

简介 Introduction:

🎉An embedded package for Florence-2 for quick interrogator. This is an attempt at a project written using GPT, which simply integrates the tagging functionality of the Florence model and adds a GUI. It mainly implements a one-click activation for tagging. Users can select a folder and set the complexity of the tagging. The feature to choose a custom model has been added, allowing users to define their own tagging terms, facilitating the use of third-party fine-tuned models.

使用方法 How to use:

让我康康用来浏览目录,或者直接复制目录地址到提示框 Let me browse the directory, or directly copy the directory address to the prompt box.

使用打标得到简单的结果 Use "tagging" to obtain simple results.

使用给我用力打标得到详细的结果 Use the feature to "tag forcefully" to obtain detailed results.

使用像唐僧一样打标得到啰嗦的结果 Tag like Tangseng to get verbose results.

使用自定义,并填入自定义词来使用第三方微调模型,如果有自定义打标触发词的话 Use custom settings and enter custom trigger words to use third-party fine-tuned models, if there are custom tagging trigger words.

将需要增加的Florence-2模型的目录放入**_internal\model‸**的目录下,刷新后就能看到,选择使用 Place the directory of the Florence-2 model to be added into the _internal\model directory, refresh to see it, then select to use.

覆盖可以覆盖有标签的文件,当然也能跳过 Overwrite can cover files with labels, or it can be skipped.

现在支持多batch打标了,4090 24batches可以做到简单打标最快0.12秒一张,复杂打标也只需要0.36秒一张,可谓极速!

处理子文件夹将迭代文件夹下所有子文件夹的图片 Processing subfolders will iterate through the images in all subfolders under the folder. image

Please download the complete package from here: Baidu Pan: https://pan.baidu.com/s/1BMEd6vG7KuDAGwVlrGk91g?pwd=36n7 or Quark Pan: https://pan.quark.cn/s/eb36e4d0a4e1 code: X4qU 使用百度或者夸克盘下载

You can download the updated main exe file to update from the release 每次的主程序更新都在release里可以下载到