Train model

[mark2146]

Hello, sorry to bother you. I followed your process once. At present, I can successfully compile tiny_yolo and it is successful. However, I now replace tiny_yolo with the traffic signal I want and then quantify the process after the debate. Everything went smoothly, but when I ran the version, I encountered an error. He said that the .so file of my model was not found. I checked a lot of information on the Internet, but most of them were ZCU versions. It seems that pynqz2 does not have it. I want to ask where is the problem?

[wutianze]

Hello, sorry to bother you. I followed your process once. At present, I can successfully compile tiny_yolo and it is successful. However, I now replace tiny_yolo with the traffic signal I want and then quantify the process after the debate. Everything went smoothly, but when I ran the version, I encountered an error. He said that the .so file of my model was not found. I checked a lot of information on the Internet, but most of them were ZCU versions. It seems that pynqz2 does not have it. I want to ask where is the problem?

It seems that your libdpumodeltiny_yolo.so is missing(/usr/lib, /lib). Pls try the mnist_tf demo first to see if it works

[mark2146]

I try it is work.So I don't know what problem?

[wutianze]

I try it is work.So I don't know what problem?

Maybe you could edit the file " runYolo.cpp" and change the kernel info to yours (like #define YOLOKERNEL "testModel"). And I recommend you try the initial model first before using your own model. And this is a more detailed tutorial about how to run yolo in pynq-z2: https://github.com/andre1araujo/YOLO-on-PYNQ-Z2. Hope this could help you.

[mark2146]

Yes,I try the initial model is successful,but try own model is unsuccessful.But I only change the model,other I don't change.