teng

It seems that there is something wrong with the conv layer. Try using the latest code conversion and Library of TNN. Or have you made any other changes?

See which layer conv0 is and whether the parameters are correct. Let's see under what circumstances this error will be reported in the initialization part of the TNN source code.

没看懂，什么叫把自己的cmake过的算法塞进去？

先看懂demo的流程，看懂再尝试修改。不要急。

需要会安卓开发，不会的话就先放放吧。

这么问那对你就是挺复杂的

"model layer output name", stride, { anchors },

> Also, in the official yolo demo of TNN, https://github.com/Tencent/TNN/examples there's no such parameters to assign the layer information. I wonder have you done any processes in it? The official...

https://github.com/Tencent/TNN/blob/0278cdbfd7480b471c6a7492e6deedfa5d48229d/examples/base/object_detector_yolo.h#L62-L71 https://github.com/Tencent/TNN/blob/0278cdbfd7480b471c6a7492e6deedfa5d48229d/examples/base/object_detector_yolo.cc#L49-L57

https://github.com/Tencent/TNN/releases Download and replace with the above address. The model can go to the official website to see the provided examples and try to figure it out by yourself.