EddieEduardo

icon company Harbin Institute of Technology icon location Harbin icon location Ph.D. candidate, computer vision (industrial defect detection currently)

Results 12 comments of EddieEduardo

Cannot achieve the comparable accuracy results with the provided

Plus, my PyTorch version is 1.8.0.

No module named 'dcn_v2'

请问这个问题最后怎么解决的呀？

Non-Max-Suppression

Thanks，I mistook it as the first score value of a tensor.

train YOLO with self data（使用 YOLO训练自己的数据）

请问收敛吗？

CUDA memory error

@wangaixue OH, JESUS!!! Mine is just 4G , are there any good solutions making it compatible with PC？Thanks for your reply.

Version 1

Hi, thanks for your reply, could you provide a link?

Version 1

Hi, I appreciate it very much! Have a nice day!

COCO Dataset

Hi, when directly training TransTrack on MOT without pretrained model (crowdhuman_final.pth) for 30 epochs, seemes it still cannot correctly localize and boundingbox **any objects** in the frame. Is it because...

ModuleNotFoundError: No module named 'layers.reorg._ext.reorg_layer._reorg_layer'

> Did you run `sh make.sh` > Did you run `sh make.sh` first? Hi,how are you? You mean that if I want to directly run train.py I also have to...

depthwise_conv2d_implicit_gemm slower than nn.Conv2d

Hi, I encountered with the same problem. When using nn.Conv2d, the running time of the model is just ~0.5s, while using the DepthWiseConv2dImplicitGEMM, the time is ~6s. The batchsize is...