I encountered this error using the original version. Can this error be resolved?

[ZJDATY]

[ERROR:0@3.422] global onnx_importer.cpp:1064 cv::dnn::dnn4_v20230620::ONNXImporter::handleNode DNN/ONNX: ERROR during processing node with 2 inputs and 1 outputs: [Expand]:(onnx_node!/models.33/Expand) from domain='ai.onnx'

[zihaomu]

If you met an error with original opencv, please issue at opencv. The opencv_ort is experimental version. Currently, I only test it with Mac M1 chip, and it supports almost every ONNX model. You are welcome to try it.

[ZJDATY]

@zihaomu This version has solved my problem, thank you very much. It did not report any errors during the model loading and inference stages.But I encountered some situations where the original dnn functions could not be used during use. For example:

cv::dnn::readNetFromDarknet();
net.setPreferableBackend(cv::dnn::DNN_BACKEND_INFERENCE_ENGINE);
cv::dnn::NMSBoxes();

And there may be some functions that I haven't used.

I want to know whether these functions are deliberately discarded, or there are some conflicts temporarily, or there are other alternatives.

[zihaomu]

The plan of this project is the following: Totally remove the old opencv dnn, and replaced it with ORT. Since the original dnn has support onnx but only a small subset, we want to do it in a different way. And in order to make developing works clear, we decide to remove all of them.

So, it is deliberately discarded. Technically, we only support ONNX model under this version. Actually, almost all darknet models could find the corresponding onnx version.

[ZJDATY]

The plan of this project is the following: Totally remove the old opencv dnn, and replaced it with ORT. Since the original dnn has support onnx but only a small subset, we want to do it in a different way. And in order to make developing works clear, we decide to remove all of them.

So, it is deliberately discarded. Technically, we only support ONNX model under this version. Actually, almost all darknet models could find the corresponding onnx version.

Okay!