R2CNN-Plus-Plus_Tensorflow icon indicating copy to clipboard operation
R2CNN-Plus-Plus_Tensorflow copied to clipboard

This is a tensorflow implementation of R2CNN++: Multi-Dimensional Attention Based Rotation Invariant Detector with Robust Anchor Strategy.

Results 28 R2CNN-Plus-Plus_Tensorflow issues
Sort by recently updated
recently updated
newest added

Hi dear developer, I was trying to run 'train.py' but received an error-'No module named 'libs.box_utils.cython_utils.cython_bbox' when trying to import a 'bbox_overlaps' file. I have checked the folder and there...

你好,我训练的是自己数据集,图像大小参差,有大小为6000*4000,于是我修改train_crop的图像大小为6000,希望这样不会裁剪的太碎,削弱了框选的特征,我也没有大改什么文件参数,但是训练的时候,loss就是不收敛,并且预测的图像中有密集的绿框,但它们并没有正确框选。 请问您能提供一些思路吗?十分感谢!

我发现您的代码里面resnet.py中将senet注意力模块以及inception模块给注释掉了。 采用自己数据集 我加回来之后发现损失变为nan 您是将这两个模块写在了其他的地方吗

I am using the default configuration file and getting multiple detections on the same object. NMS is not working properly. Can you tell me why ?

Please ask the author, why my visualization result is so bad? ![image](https://user-images.githubusercontent.com/49429987/105676570-aa44db00-5f25-11eb-8281-bcdca7186585.png) ![image](https://user-images.githubusercontent.com/49429987/105676585-b29d1600-5f25-11eb-9a66-e4f94c1f084f.png)

你好!我想问这个代码是否不包含iou-smoothl1LOss函数的代码,我用一块1080的GPU训练33W步对这个repo复现,始终得到的map只有0.60,对比你发的2019ICCV的SCRDet中的0.72还是有着较大的差距 ,按道理没什么改动不应该有这么大的出入。我想请教一下这其中是哪里有问题 还是这个repo只是部分代码要我自己去整合其它的代码块?麻烦给出一下回复,谢谢你!

在demo_rh_pyramid.py中的157行如下面代码: det_boxes_r_ = forward_convert(det_boxes_r_, False) det_boxes_r_[:, 0::2] *= (w_len / ss) det_boxes_r_[:, 1::2] *= (h_len / ss) det_boxes_r_ = back_forward_convert(det_boxes_r_, False) 有时会报IndexError: too many indices for array,查看有时为空。当用if做排除后还会出现此错误,请指教一下。如果方便的话可以给下微信?谢谢了

good first issue

Hi, I'm trying to run the demo code and I followed all the step that you mentioned in the Markdown. But I keep on getting this error ``` python demo_rh.py...

why does run "python setup.py build_ext --inplace" compile /libs/box_utils and /libs/box_utils/cython_utils have error? I have tried R2CNN-PLUS-PLUS, R2CNN_Faster-RCNN and R2CNN_FPN, all of them have the same error!! Details is as...