PaddleOCR/latest/ppocr/model_train/detection #14232
giscus[bot]
bot
Replies: 1 comment 4 replies
-
|
请问模型导出后没有.pdmodel文件和.pdiparams.info文件，反而多了inference.json文件是什么原因呢，模型导出命令中使用的就是训练用的配置文件和训练之后的best_accuracy |
Beta Was this translation helpful? Give feedback.
-
|
@laochi-love-cmy1314 这个是新的pir导出格式，如果想回到旧版本可以使用 <= paddle3.0 beta 1.0 版本的paddle。也可以直接使用新版本的模型。 |
Beta Was this translation helpful? Give feedback.
-
|
请问如何使用最新版本呢？我使用PaddleOCR（lang="ch",det_model_dir="",cls_model_dir="",rec_model_dir=）这个函数，设置了三个模型目录，按官网之前的那种是可以的，但是我把目录换成这种新导出的模型目录去识别的时候，会先下载缺失的.pdmodel文件和.pdiparams.info文件，这种应该不是使用我自己训练的模型吧 |
Beta Was this translation helpful? Give feedback.
-
|
可以先切换到旧版本，新版本可能还在适配中。 https://www.paddlepaddle.org.cn/install/quick?docurl=/documentation/docs/zh/develop/install/pip/linux-pip.html 根据自己的需要按照对应的paddle版本然后 把 3.0.0b2 改成3.0.0b1 |
Beta Was this translation helpful? Give feedback.
-
|
好的哥，非常感谢 |
Beta Was this translation helpful? Give feedback.
-
PaddleOCR/latest/ppocr/model_train/detection
Awesome multilingual OCR toolkits based on PaddlePaddle (practical ultra lightweight OCR system, support 80+ languages recognition, provide data annotation and synthesis tools, support training and deployment among server, mobile, embedded and IoT devices)
https://paddlepaddle.github.io/PaddleOCR/latest/ppocr/model_train/detection.html
Beta Was this translation helpful? Give feedback.
All reactions