DebeshJha/Kvasir-Instrument

Kvasir-Instrument: Diagnostic and therapeutic tool segmentation dataset in gastrointestinal endoscopy contains the information about the Kvasir-Instrument dataset

🔍 Use-Cases:

Automatic surgical instrument segmentation
Data augmentation studies
Benchmarking of the different deep learning methods
Transfer learning
Model interpretability studies

Dataset information:

The Kvasir-Instrument dataset includes 590 frames consisting of various GI endoscopy tools used during both endoscopic surveillance and therapeutic or surgical procedures. Moreover, information about the dataset uses, their application, annotation protocol can be found from their webpage.

Downloadable link:

The dataset can be downloaded from https://datasets.simula.no/kvasir-instrument/.

Research and publication

Arxiv

OSF preprint

Original publication

Evaluation metrics

We suggest evaluation metrics such as dice coefficient, Jaccard index, recall, precision, and F2.
Their description can be found here.

## Results on challenging images The figure shows the results on the challenging images.

Citation

Please cite our paper if you find the work useful:

@inproceedings{jha2021kvasir,
  title={Kvasir-instrument: Diagnostic and therapeutic tool segmentation dataset in gastrointestinal endoscopy},
  author={Jha, Debesh and Ali, Sharib and Emanuelsen, Krister and Hicks, Steven A and Thambawita, Vajira and Garcia-Ceja, Enrique and Riegler, Michael A and Lange, Thomas de and Schmidt, Peter T and Johansen, H{\aa}vard D and others},
  booktitle={Proceedings of the International Conference on Multimedia Modeling},
  pages={218--229},
  year={2021},
  organization={Springer}}

Contact

please contact debesh.jha@northwestern.edu for any further questions.