Replies: 2 comments
-
可以参考:https://github.com/search?q=repo%3Ajingsongliujing%2FOnnxOCR%20shape&type=code |
Beta Was this translation helpful? Give feedback.
-
在使用导出的ONNX模型通过TensorRT进行推理时,设置输入Tensor的动态形状(Dynamic Shape)是一个关键步骤,因为它直接影响推理的性能和适配性。对于PP-OCRv3的 1.
|
Beta Was this translation helpful? Give feedback.
-
det模型: en_PP-OCRv3_det_infer.onnx
rec模型: en_PP-OCRv3_rec_infer.onnx
我们知道用TensorRT来队PPOCR的模型进行推理需要设置的输入Tensor(X)的Dynamic Shape,那这个Dynamic Shape该设置为多少合适?我的意思是det模型和rec模型的min/opt/max 三个shape参数要如何设置?谢谢!
Beta Was this translation helpful? Give feedback.
All reactions