EdSong
EdSong
> 您好,我也没做过目标检测的任务。目前我有个downstream task用到了。 在使用阶段,可以输出map值, 但是没有绘制了检测框的图像输出,我想问问您关于这一块儿是需要怎么处理,我在augmentations里面看见了imgshow_output 方法,但是会发生错误。 不是很清楚你的问题,请问你是不想要检测框只想要热力图吗
In the annotation, we use relative coordinates. Thus, when you want to use the annotation, you can multiply them by the width and height of the image. Of course, if...
Yes. But you better resize the target images to the same level as the source images.
Actually, this is what the paper talks about. If you model is trained on one type and evaluated on another type, it will definitely cause overfitting. So, ori+type1 will overfit...
It is obvious that ori+type1 can improve the mAP in ori. The training and the evaluation are in the same domain (both training data and test data contain ori). type1...
Are the training size and the test size the same? If not, there will be a performance drop. Your WQT output size is 600, so you better not use a...
我猜你应该是一个中国人,我用中文回复好了,清晰一点。 1. 这种性能的提升不遵循线性关系。 2. 域泛化问题,本质上还是一个泛化问题(Generalization)。那个大表格的意思是想表达,我们测试域的性能和我们训练域与测试域之间的距离是有关系的,训练域和测试域越接近,那么测试域的性能就越高。但是,我们不可能能够知道未来在实际应用的时候,真正的测试域是什么样的(type8),而且仅仅是简单地把训练域的数量增加,也并不能完全解决这个问题(type8上30map的性能和同域训练测试的56+map性能还是有很大差距)。说明我们还需要一些别的技巧以让模型尽可能学习到域无关信息。当然DG-YOLO提出的模块确实比单纯用WQT带来了提升,然而这个结论也说了仅仅这样还是不够的,这个问题还需要进一步去研究。 同学可以参考我的最新工作,希望能够给你带来启发:https://github.com/mousecpn/DMC-Domain-Generalization-for-Underwater-Object-Detection
do you have any GPU?
modify the code like that: cont_feat = net.encode_with_intermediate(low_cont.unsqueeze(0)) style_feat = net.encode_with_intermediate(low_style.unsqueeze(0)) coeffs, output = model(cont_img.unsqueeze(0), cont_feat, style_feat)
那应该是是复现的问题,改成range(1,4)把