Mike

Moving data from `cpu` to `gpu` is very expensive. Some users will most certainly want the output in `torch.cuda` format for further processing on `gpu`. If you move data to...

Good experiment! But you have to do this for all the model outputs: ```pyhton torch.from_numpy(prediction.bboxes_xyxy).cuda() torch.from_numpy(prediction.confidence).cuda() torch.from_numpy(prediction.labels).cuda() ``` If you then what to further concat them you may get a...

Sure, I have some valuable feedback, so let's chat :smile: ! Send me a meeting invite at [[email protected]](mailto:[email protected])

I fixed some bugs. It is stable now.

> We can't include the code in our repo, since unfortunately, it is under AGPL 3.0 license (our repo is Apache 2.0) Yes, I see > But we will definitely...

Deci-AI cannot provide NPU model versions for you. This is because you need to compile a program for the specific chip you run your models on. Each of them have...

Same here but with `additional_targets` instead of `ReplayCompose` which should be very similar

This: https://github.com/albumentations-team/albumentations/issues/716#issuecomment-760723917 didn't fix it for me

Same problem here as @TUDelftHao  Just for comparison. I let YOLOv5 with pretrained weights train one epoch on my custom dataset and got 68AP

👀