lzylzylzy123456

orin nx是有输出的，是不是MNN版本太低了。

期待TensorRT版本，我有一个小疑问，请问预估能在Nvidia什么型号上的边缘设备进行25ms的推理呢？Orin Nx、Xavier Nx、Orin Nano?最后再次感谢你们精彩的工作！

I have encountered this problem. Have you encountered it before？ Traceback (most recent call last): File "main_open_world.py", line 366, in main(args) File "main_open_world.py", line 290, in main model, criterion, data_loader_train,...

> Yes, that's the case! An image may contain only instances that do not belong to any valid classes during one specific task, then all the instances (annotations) in the...

> Have you successfully reproduced OW-DETR before? I successfully ran through OWOD, but I was unable to run through OW-DETR, as the issuer stated, in open_world.py seems to have triggered...

  The first image shows the task classification of OW-DETR, and the second image shows the task classification of OWOD. After replacing the task classification of OW-DETR with the...

> >   The first image shows the task classification of OW-DETR, and the second image shows the task classification of OWOD. After replacing the task classification of OW-DETR...

> >   The first image shows the task classification of OW-DETR, and the second image shows the task classification of OWOD. After replacing the task classification of OW-DETR...

What I mean by segmentation here is data segmentation

> In OW-DETR, the authors declared that they used the replay mathod to prevent the model from catastrophically forgetting what it have learnt, but I cannot find the corresponding codes...