-
Notifications
You must be signed in to change notification settings - Fork 115
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
Iteration Plan for OBBDetection 2021-11-19 #72
Comments
Looking forward to the implementation of GWD and KLD in this repo. |
你好,我正在基于OBBDetection移植s2anet,代码在s2anet,但是在12个epoch下精度目前卡在72.7。查看原始版本的s2anet的训练log发现似乎数据集切分后的图像数量有区别,这边batchsize8的时候每Epoch1600步,s2a里是2631步,我准备再尝试一下2x调度下精度能否正常。您能帮忙看看是哪里出了问题吗,感谢。 |
好的,我周末测试一下,尽量对齐精度,整理一下代码再提pr |
@liuyanyi 您好,我今天仔细地看了一下,发现造成精度不对齐也许不是裁图问题,s2anet在训练地时候应该也将没有目标地图像块过滤掉。 主要问题还是在于对有向目标地定义方面,OBBDetection定义有向目标逆时针旋转方向为正,而S2Anet中定义顺时针方向为正,这导致在coder、IoU calculator、alignconv处都有一定地变换。 我发现您地代码中coder使用的还是S2ANet的,这与OBBDetection传入的gt_obboxes定义的角度有出入,并且,alignconv处的xr,yr计算公式也需要发生变化。 |
哦哦,之前没注意过iou calculator和Align Conv,我取看看,感谢 |
果然是角度的问题,调了角度之后在OBB的角度定义下可以达到74.0,已经提了一个pr。 |
Iteration Plan for OBBDetection 2021-11-19
BboxToolkit:
OBBDetection
It's very welcome to propose PRs.
The text was updated successfully, but these errors were encountered: