Object Pose Annotation Tool

https://autognc.github.io/opat-js

A lightweight interactive 6D pose annotation tool. Loads images and overlays a 3D model on top of them, allowing the user to line up the model and an object in the image using keyboard shortcuts.

Usage

Loading a model

Currently, the only supported model format is a single .glb file which is loaded using the three.js GLTFLoader. The model is retrieved from gltf.scenes[0].children[0]; any additional models or scene data is ignored.

Loading intrinsics

Once images and a model have been loaded, camera intrinsics can be loaded from a JSON file. There must be an entry for every image with the format:

{
    "image_file_name.png": {"fov_y": vertical_fov_in_degrees}
}

Currently, the horizontal FOV is calculated from the aspect ratio of the image and there is no support for differing vertical/horizontal pixel lengths. The principal point is assumed to be the center of the image; there is not yet support for a crop or offset.

Keybindings

See src/config.js for keybindings. When the position and rotation readout at the top left is green, that means the current pose is saved.

Development

This project was created with Create React App.

Enter the project directory and run npm install. To run the app locally, run npm start. To deploy to GitHub pages, run npm run deploy.

Docker

Run the docker container by running:

env PORT=[local opat-js port] docker-compose -f opatjs_compose.yml up