Pose Estimation & Semantic Segmentation Annotator Models

[jdp8]

Description

The changes include adding the Pose Estimation & Semantic Segmentation Annotator models along with the pre-processing functions required to get the ControlNet input image which is done using OpenCV.js.

Specific Changes

• Add examples/react/src/annotator_helper_functions.ts file which contains necessary functions to run both models.
• Add Annotator model and Annotator config upload elements in examples/react/src/App.tsx.
• Modify uploadImage() function in examples/react/src/App.tsx to account for Pose Estimation & Semantic Segmentation.

Notes

• Pose Estimation was taken from here.
• Semantic Segmentation was taken from here.
• The Pose Estimation and Semantic Segmentation models can be downloaded from here.
• Getting the annotated image takes ~17s with Semantic Segmentation and 17-24s with Pose Estimation (depends on the selected model) on an M2 Max.
• The annotator files must be uploaded before uploading the ControlNet image.
• I'll submit an issue asking if it's possible to convert all the Annotator models to ONNX which would give us access to every ControlNet model in JavaScript.