y78h11b09 issues

Results 11 issues of


                                            y78h11b09

It does'nt work for dectecting 1600 objects for the funtion "torch.clamp(classification, min=1e-4, max=1.0 - 1e-4)" in focol_loss

Recently, I found that efficientDet-do didn't work for 1600 objects because cls_loss still didn't downgrade in training. So , i modified the funtion **"torch.clamp(classification, min=1e-4, max=1.0 - 1e-4)" into "torch.clamp(classification,...

why the last layer of retina_head is nn.con2d() of 3*3 ,not is nn.linear() , what difference are both of them??

training efficientDet-d0 over VOC0712 dataset, cpu cost is very high, always over 85%. can anybody tell me what happens ? thanks

The demo.py has too many boxes to incorrect, and the code updated is unstable.

Undefined variable of scale and the key ' img' of getitem in voc20712.py , but used both of them in the eval.py

How to compile the lib on the win10+pytorch0.4.1+cuda9.2?

Hi ,I had successufully compiled the nms of lib by 'python setup.py install', but later, i got the problem, as follows: Traceback (most recent call last): File "E:/Pytorch_Room/faster-rcnn_pytorch-master/demo.py", line 31,...

About the building of vocabulary?

Hi, can you tell me why you use the training data to build vocabulary and not use the complete data?

Can you sent me your pretrained weight?

Hi, I trained your model on mmdetection, but i have got the mAP 22.7 on EfficientDet D3 , with 2GPUs 1080ti. Can you sent me your pretrained weights ? thanks...

why your reference mAP in orginal paper is difference to 'Yet-Another-EfficientDet-Pytorch'??

Yet-Another-EfficientDet-Pytorch mAP 0.5:0.95(paper) -- D0 | efficientdet-d0.pth | 1049 | 36.20 | 163.14 | 33.1 | **33.8**

The calculation of loss in ATSSS is different from FCOS and original ATSS, why???

Thanks for your great work, and we appreciate it a lot. I train your code long time, about 300 epoches on d3, but i haven't got the better mAP than...