Repository Issue Activity (beta)

#9060 encode_jpeg generates corrupted JPEG data when using CUDA

#9594 XYXY to CXCYWH conversion overflows for narrow dtypes

#2135 Anchors not being properly distributed across feature maps?

#9577 `ElasticTransform` is slow for larger images and/or sigma

#9576 `transforms.v2.Normalize` caused small `H2D` and cpu waits gpu if process image on gpu

#9513 deform_conv2d kernels are launched on the default CUDA stream

#7959 Faster R-CNN training OOM at ops/boxes.py _box_inter_union function

#9298 ROCm Build from source hit error: multiple definition of `vision::cuda_version()'

#9560 `torchvision.transforms.v2.functional.pad` should have maintained the `channels_last` format of the input

#9542 Image decoders feature support

#9547 Make `to_image` reading return tensors in CHW

#9551 Out-of-bounds read in the GIF decoder when decoding an untrusted image

#9532 Container for multi-pass and randomized subset transform loops

#9536 datasets: CIFAR10 download slow

#9534 SuperCompress: Free open-source LLM prompt compression - cut token costs by ~65%

#9337 Make "RuntimeError: operator torchvision::nms does not exist" error better

#9519 Stray torchvision==0.27.1+df56172 (win_arm64-only) wheel on the stable /whl/cpu index breaks resolvers

#9348 The URL for the SUN397 dataset isn't working

#8875 Celeba dataset can not be downloaded

#9517 Path traversal in TorchVision dataset archive extraction

#9501 viable/strict updating silently broken

#603 Strange behavior trying to install in conda with PyTorch built from source

#5336 port datasets from the old to the new API

#5754 [RFC] `torchvision.transforms` revamp

#5626 Clean up prototype transforms before migration