Nicolas Hug comments

Results 570 comments of


                                            Nicolas Hug

[FBcode->GH] [fbcode] Reduce log spam from torchvision video reader

We can't import this diff as it's breaking the CI with a compilation error. I'll close the PR and backout the diff in D58283929

Building on archlinux fails with include error

Hi @binarynoise can you please copy/paste the logs here? It is safer for us than to download a file.

Building on archlinux fails with include error

Sorry @binarynoise it's hard to reproduce - our build jobs are running fine on our CI, so I would assume this has something to do with your env. Have you...

PyTorch standard Coco dataset (datasets.CocoDetection) not compatible with Faster R-CNN object detection model

Hi @ranjaniocl , try to print the input that gets passed to `RandomIoUCrop()`. There should be bounging boxes and PILimages/tensors in there. If not, it's likely that the pipeline is...

PyTorch standard Coco dataset (datasets.CocoDetection) not compatible with Faster R-CNN object detection model

@ranjaniocl sorry it looks like your issue might be more in scope for https://discuss.pytorch.org/

position, colour, and background colour of text labels in draw_bounding_boxes

Thanks for opening this issue @carandraug. I think the proposal is reasonable, the current position of the label does make them difficult to read. In terms of API / functionality,...

position, colour, and background colour of text labels in draw_bounding_boxes

That SGTM, thanks for the details. Provided that the licensing terms of matplotlib allow it, I hope there exist a piece of code that we can just copy/paste from matplotlib...

position, colour, and background colour of text labels in draw_bounding_boxes

> I If so, I will prepare a PR for text colour/etc first and then another for the label positioning Yes, that sounds good. Happy to consider better default for...

Generalising FakeData

Thanks for the PR @ctr26 , but what is the benefit of `FakeImageFolder` over the existing `FakeData` class?

Generalising FakeData

Thanks for the details. IIUC you just need the signature of `FakeData` to be compatible with that of `VisionDataset` in order to avoid special-casing the call to `FakeData`? Instead of...