Mission
Manual annotation and counting of entities in underwater photographs is a central element of most census studies. With exploding jellyfish populations worldwide, in-situ studies of jellyfish polyps are becoming crucial for understanding and predicting the population dynamics. However, these are limited to small sample sizes due to tedious manual labor involved in counting in underwater photographs. We have developed an automated polyp counting algorithm PoCo, which allows processing orders of magnitude greater collections of images than previous possible.
PoCo-v2.0
PoCo-v2.0 is the most recent and easy-to-use automated polyp counter. We also provide an annotation tool that will allow the researchers to quickly adapt and train PoCo on their counting problems.
Try it out
- A Python source code for PoCo-v2.0 tool is available at GIT here.
- Windows pre-compiled application available here (Tested on Windows 10, Nvidia GTX 1070).
- See quick tutorial on how to run PoCo here.
Datasets
- PoCo-v2.0 was trained on a large publicly available polyp dataset.
Previosus versions
A preliminary version, PoCo_v1.0, is available here.
Citing PoCo
If you use PoCo-v2.0 in your research, please cite the following paper:
@article{ZavrtanikSeg2020,
title = "A segmentation-based approach for polyp counting in the wild",
journal = "Engineering Applications of Artificial Intelligence",
volume = "88",
pages = "103399",
year = "2020",
issn = "0952-1976",
doi = "https://doi.org/10.1016/j.engappai.2019.103399",
url = "http://www.sciencedirect.com/science/article/pii/S095219761930315X",
author = "Vitjan Zavrtanik and Martin Vodopivec and Matej Kristan"
}