Torchvision Transforms

preview_player
Показать описание
Philip Meier from Quansight presents "Torchvision Transforms" at PyTorch Conference 2022.

TorchVision is extending its Transforms API! This talk previews the current prototype that is no longer limited to Image classification, but can also natively handle Object Detection, Instance and Semantic Segmentation, and Video classification.

#PyTorch #ArtificialIntelligence #MachineLearning
Рекомендации по теме
Комментарии
Автор

Will the new API support polygons or oriented bounding boxes (bbox with angle)?

nullkothenature-lovingbear
Автор

Hello! Do I understand correctly that it is not possible to use touchvision.transforms methods for 1d tensors?

maryamshangaray