Instance Segmentation Metrics

Instanseg is an open-source python library that provides various metrics for evaluating the results of the algorithms for segmenting and associating instances.

List of metrics implemented in the library:

• Summary segmentation metrics
  • Panoptic
  • Full statistics
• Instance-based segmentation metrics
  • Precision
  • Recall
  • F-Score
  • Under segmented ratio
  • Over segmented ratio
  • Noise ratio
  • Missed ratio
• Point-based segmentation metrics
  • Intersection over Union (IoU)
  • Dice
  • Mean of some metric for matched instances

For more, please visit the Instanseg documentation.

You can also find full information about the project on the Instanseg project website.

Python quick start

Library can be installed using the pip package manager:

$ # Install package
$ pip install instanseg

$ # Check installed version of package
$ pip show instanseg

Example of usage

Below is an example of using the precision metric:

>>> from instanseg.metrics import precision
>>> pred_labels = np.array([1, 1, 3, 3])
>>> gt_labels = np.array([2, 2, 0, 3])
>>> tp_condition = "iou"
>>> precision(pred_labels, gt_labels, tp_condition)
0.5

Citation

@misc{kornilova2022evops,
      title={EVOPS Benchmark: Evaluation of Plane Segmentation from RGBD and LiDAR Data}, 
      author={Anastasiia Kornilova, Dmitrii Iarosh, Denis Kukushkin, Nikolai Goncharov, Pavel Mokeev, Arthur Saliou, Gonzalo Ferrer},
      year={2022},
      eprint={2204.05799},
      archivePrefix={arXiv},
      primaryClass={cs.CV}
}

License

This project is licensed under the Apache License - see the LICENSE file for details.