92ypli

Results 11 comments of 92ypli

Hi ， excellent work！ In your paper ,you said the FCOS can replace the RPN in two-stage framework.So how shoulf I change the codes? Could you share your opinions?

> > > @zsl1996 1) when using FCOS in Faster R-CNN, I think the original alternative 4-step training strategy should be used instead of the end-to-end training because anchor-boxes can...

Why did my test only achieve mAP: 0.409?

Same here. How to fix it?

how to understand this:loss_conf = self.obj_scale * loss_conf_obj + self.noobj_scale * loss_conf_noobj

why self.noobj_scale = 100？

General learning VOC how many epochs convergence?

How much is the mAP?

operation requirements

shuffle op

operation requirements

Not Supported OP: shufflenet Framework: pytorch or onnx

operation requirements

Sorry, my meaning is ShuffleChannel.

operation requirements

Framework: onnx Not Supported OP: Transpose

Question about ground truth data?

> Hi, My gt data has four channels, they are x1 channel, y1 channel, x2 channel and y2 channel, so my gt data is 4 x width x height, but...

DukeMTMCT数据问题咨询

请问你们还有数据集吗