DMZ Crane Watch

DMZ Crane Watch is a new and ongoing project. We have two intertwined goals. First, the result will be used for ecological studies of the cranes in and around the DMZ. With your clicks, we can learn more about how cranes make use of the rice fields next to the DMZ. We collect a range of information – species, numbers, behaviors, time/date, and temperature – to better understand the spatial and temporal distribution of the cranes and their interactions with one another. We are especially keen to examine the ways in which cranes respond to the increased number of infrastructures (roads, greenhouses, buildings) in relation to the recent modification in the delimitation of the Civilian Control Zone; the demarcation line was relocated in the northern part of Cheorwon. This data will allow us to identify the important changes in the quality and quantity of the crane habitats, critical to our conservation efforts, and to consider the salience of rice fields as wintering habitats for these endangered birds.

AI provides species-specific crane numbers by analyzing the photograph.

Second, the annotated data set will be used to improve the AI that can count the number of the cranes by their species. We have already developed crane-counting AI by making use of photographs produced by ecologists (Fig 5). The result will be published at 2021 IEEE International Conference on Image Processing in September 2021. However, we learned that the machine trained with human-produced images needs lots of improvement to work with remote-sensing images. There are major differences between ecologists’ photos and trail cam images. That’s why we are asking your help to create a new dataset with trail cam images.

The engineering team working on the AI part of this project has published one article, hereafter:
H. Go, J. Byun, B. Park, M. -A. Choi, S. Yoo and C. Kim, "Fine-Grained Multi-Class Object Counting," 2021 IEEE International Conference on Image Processing (ICIP), 2021, pp. 509-513, doi: 10.1109/ICIP42928.2021.9506384.