A Tensor Image is a tensor with (C, H, W) shape, where C is a The transformations that accept tensor images also accept batches of tensor PIL images, or for converting dtypes and ranges. The Conversion may be used to convert to and from
Most transformations accept both PIL imagesĪnd tensor images, although some transformations are PIL-only and some are This is useful if you have to build a more complex transformation pipeline Transforms give fine-grained control over the Most transform classes have a function equivalent: functional Transforms are common image transformations available in the More about the APIs that we suspect might involve future changes. Please submit any feedback you may have here, and you can also check Note that these transforms are still BETA, and while we don’t expect majorīreaking changes in the future, some APIs may still change according to userįeedback. Transforms v2: End-to-end object detection example. These transformsĪre fully backward compatible with the current ones, and you’ll see themĭocumented below with a v2. Not just images but also bounding boxes, masks, or videos. 2 namespace, which add support for transforming In 0.15, we released a new set of transforms available in the
