s2anet

s2anet copied to clipboard

OSError: work_dirs/epoch_12.pth is not a checkpoint file

3

I trained the model on dota, but when I try to test it it throws this error. Traceback (most recent call last): File "tools/test.py", line 236, in main() File "tools/test.py",...

rohansig

训练完的模型大小（290.8MB）与发布的模型大小（160MB）不一致，是什么原因导致的？

1

使用单卡训练DOTA数据集的一部分，训练完后的保存的模型大小，与作者提供的模型大小不一致。

xc-chengdu

使用dota1.5时，container crane的txt文件显示为空

1

请问一下，把dota1.0换成Dota1.5之后需要更改哪些配置？为什么我用dota1.5时最后container crane的txt文件显示为空？是数据集的问题，还是因为我的代码没有改全？

yx618

Excuse me, have you ever train s2anet on UCAS_AOD dataset? I have problems aligning the accuracy in [UCAS_AOD_Benchmark](https://github.com/ming71/UCAS-AOD-benchmark) which you mentioned in another issue. And my model have a low...

jedibobo

Unused conv in or_pool

1

Hello, thanks for the great work. In the RotationInvariantPooling module, a convolution + batchnorm is being initialized but seemingly not used later on. Is this for testing purposes? Can both...

danielbsf

Environment for pytorch1.9 version

1

when running python setup.py develop in pytorch1.9 version using docker, with environment as python==3.8, torch==1.9.0 and cudatoolkit==11.0, there exist error: /opt/conda/envs/s2anet/compiler_compat/ld: cannot find -lc10_cuda /opt/conda/envs/s2anet/compiler_compat/ld: cannot find -ltorch_cuda collect2: error:...

YuxuanChen21

S2ANet中的对齐卷积能使用NCNN移动端部署吗

1

17602530266

图片预处理时出现错误

1

我在处理dota1.5数据集的时侯，出现了如下错误，请问一下我该如何解决？谢谢 File "/home/speech/cipan/s2anet/DOTA_devkit/convert_dota_to_mmdet.py", line 66, in convert_dota_to_mmdet bboxes, labels, bboxes_ignore, labels_ignore = parse_ann_info(label_path, img_name) File "/home/speech/cipan/s2anet/DOTA_devkit/convert_dota_to_mmdet.py", line 32, in parse_ann_info labels.append(label_ids[class_name]) KeyError: 'container-crane'

yx618

关于训练epoch数

3

请问对于dota数据集训练s2anet来说，一般训练12个epoch就够了么？还是说需要训练上百甚至几百个epoch。实际上我在训练自己的遥感数据集时，loss在epoch=120、200时，loss大概能从0.08降到0.06左右，但是mAP相差不多，甚至更低了，请问一般应该如何选择训练epoch数呢？ 另外，请问对于类别分布不均的情况，有的几千有的只有几十个，可以使用什么方法解决呢？谢谢！

galaxyGGG

请问，训练的时候一直卡在这一部是怎么回事

4

一直卡在下面的部分，显存只占用600m  强制中断后的错误如下 

tudoulei