Philip Meier
Philip Meier
This is happening since roughly a month (first detection in #6274), but was masked by #6268. http://www.cs.virginia.edu/~vicente/sbucaptions/ The server is still up, but the path is no longer valid. I'll...
Adresses #6268. The bot is spamming this once per day. Let's revisit this before the next release to see if we can reinstate the test or disable downloads for Places365...
In our detection references we have a couple of transformations that can handle images and bounding boxes simultaneously: - [x] [`RandomPhotometricDistort`](https://github.com/pytorch/vision/blob/e836b3d8baf8cbfdd18fcc872631962345a65689/references/detection/transforms.py#L227) #5663 - [x] [`RandomZoomOut`](https://github.com/pytorch/vision/blob/e836b3d8baf8cbfdd18fcc872631962345a65689/references/detection/transforms.py#L164) #5551 - [ ] [`RandomIoUCrop`](https://github.com/pytorch/vision/blob/e836b3d8baf8cbfdd18fcc872631962345a65689/references/detection/transforms.py#L77)...
This expands the prototype test matrix to test all Python versions on Linux. This should give us some extra safety to catch stuff like #5801 and https://github.com/pytorch/data/runs/6500848588#step:9:1977, which are not...
When debugging vision models, it is often useful to be able to map predicted bounding boxes, segmentation masks, or keypoints back onto the original image. To do this conveniently, each...
Right now if you have @sourcery-ai-bot installed for your GitHub repository it opens a separate PR. In this new PR it simply adds all refactorings without comments. If you let...
Consider the following structure of my package: ``` . ├── bar │ ├── baz.py │ └── __init__.py ├── foo │ ├── baz.py │ └── __init__.py └── __init__.py ``` In both...
I've tried to tackle static typing and got a vendorable protocol that can be checked statically as well as at runtime for all but one case I'm going to detail...
Due to the limitations of floating point arithmetic, comparing floating point values for bitwise equality is only required in very few situations. In usual sitatuations, for example comparing the output...
The repr has shape `BxPxCxHxW`. The current method cats along the batch dimension `B`: https://github.com/pystiche/pystiche/blob/71217c24557dfba05da5795547bf6f3034e7c66f/pystiche/loss/_comparison.py#L380 It should cat along the patch dimension `P`.