IPoseExtractor<T>

Interfaces · AiDotNet.Diffusion.Preprocessing

Pluggable keypoint extractor interface. Concrete implementations wrap pretrained pose-estimation networks (OpenPose / DWPose / RTMPose) and must return a [batch, 3, H, W] tensor whose channels encode the rendered pose skeleton (or per-keypoint heatmaps stacked as RGB). Plug into OpenPosePreprocessor via its constructor.

Methods

Method Summary
Tensor<T> ExtractKeypoints(Tensor<T> image) Extracts pose keypoints from an input image batch and renders them as a 3-channel skeleton tensor consumable by ControlNet’s pose conditioning branch.