OCR耗时问题 #16577
Unanswered
Acheerfulloser
asked this question in
Q&A
OCR耗时问题
#16577
Replies: 1 comment 2 replies
-
|
可能转onnx算子某些地方有坑吧,如果想要将paddle模型转onnx,可以参考paddleocr的官方文档 https://www.paddleocr.ai/main/version3.x/deployment/obtaining_onnx_models.html |
Beta Was this translation helpful? Give feedback.
2 replies
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Uh oh!
There was an error while loading. Please reload this page.
-
使用官方的开源识别模型en_PP-OCRv4_rec_train转onnx,在CPU上跑耗时3-4ms,使用configs/rec/PP-OCRv4/en_PP-OCRv4_rec.yml,相同框架,训练模型转onnx耗时基本上是开源模型的两倍。
只保留前半部分conv+bn+act(hardswish),两个模型的耗时,也基本上呈现1:1.5的比例。
只保留前半部分conv+bn,两个模型耗时,近似,但后者略高。
以上实验均在同一张图,同一台机器,重复测试。
请问,这个耗时问题,还有什么办法确认排查吗?或者转模型的时候有什么要注意的地方?
Beta Was this translation helpful? Give feedback.
All reactions