SPTSv2 icon indicating copy to clipboard operation
SPTSv2 copied to clipboard

The official implementation of SPTS v2: Single-Point Text Spotting

Results 15 SPTSv2 issues
Sort by recently updated
recently updated
newest added

你的数据集json文件中anotation字段里的rec": [35, 33, 44, 41, 38, 47, 50, 46, 41, 33, 96, 96, 96, 96, 96, 96, 96, 96, 96, 96, 96, 96, 96, 96, 96],这25个数字咋来的呢?后面重复的96是填充pad?这个rec字段我发现用现有的标注工具并不能生成,是要自己获取数据,然后插入rec字段么?

参数表中的chars字段展示了识别的字典,比如35相当于是字典中的第36个字符,96则是padding。可以根据字典进行映射,也可以由你自己来决定这个字典 _Originally posted by @zhangjx123 in https://github.com/bytedance/SPTSv2/issues/12#issuecomment-1874916470_

The owner of this work has left ByteDance, for more communication and cooperation, please contact [email protected]

能否写一个对应的文档介绍下自定义数据集的训练全流程以及onnx模型的导出和Tensorrt部署呢