hua-rs 0.1.0

Heirarchical Uncertainty Aggregation (Park et al., 2022) in Rust.

Hierarchical Uncertainty Aggregation

This repo implements hierarchical uncertainty aggregation from the paper Active Learning for Object Detection with Evidential Deep Learning and Hierarchical Uncertainty Aggregation by Park et al. (2022).

It is in written in Rust with bindings for Python.

Build

Run maturin build to build the wheel.

Usage

1. Import the package with import hua.

2. Create a hua.HUA object with iou_threshold and score_threshold.

3. Call hua.HUA.run() with:

   bounding_boxes: Array of bounding boxes in the format [x1, y1, x2, y2] of shape (N, 4),
   class_probabilities: Array of class probabilities of shape (N, C),
   uncertainty_scores: Array of instance level uncertainty scores of shape (N,),
   scales: Array of FPN scales from which the corresponding prediction was obtained of shape (N,).
   

4. The method returns the informativeness score of the image and the indices of the groups created by HUA.

Acknowledgements

Powerboxes: It was used to speed up the IOU calculation in the code.