康行天下

Check if the object labels are all correct, and whether the initial weight file match the proto file(for example, the proto maybe merge BN layers to convs).

It's hard to figure out the exact problem according to your information. Please check again the params setting or the proto definition or any other things.

It is ok to train with the `rfcn-*` proto, I have tested most of the proto that I upload.

Can you paste more snippet here(or in [pastebin](https://paste.ubuntu.com/)) for analysis, such as config,data label and proto.

There seems have no problem, I cannot figure it out either. You can also try with [D-X-Y's repo](https://github.com/D-X-Y/caffe-faster-rcnn/tree/dev), since my repo is based on his and changed some code.

Thanks for your question and corresponding solution. I am not familiar with the `tiny-yolo` model, can you explain the problem more concisely? By the way, the sentence `Middle-aged people have...

看了下darknet的pooling层代码, 的确如你所说, 在yolo v3之前的代码中的pooling层逻辑与caffe的不同. 但是最新的darknet代码中的pooling层又修改成与caffe一致的了. 添加一个pooling_yolo_v2名字的层是比较好,不容易混淆. 至于内存越界的问题是很难查, 哈哈..

The format is (cx, cy, w, h, θ), θ is anti-clockwise radian.

Yes, same as human normal sense.

I'm not sure either, there may be many aspect causing this error. Check your dataset's annotation, and try adjust some of the hyperparams.