Photos in social media reveal socio-cultural value of landscapes

  Summary: Every day, users upload millions of photos on platforms like Flickr, Instagram or Facebook. Now a study shows that these photos can be used to assess the social significance of some scenarios. For this purpose, he developed a new image analysis method based on artificial intelligence. Results for landscape management and presentation can be of special importance.

Working ecosystems are important for humans: they supply key resources such as food, water or fuel. But they also play a cultural role in the form of resorts, holiday destinations or ashetic photographic motifs. "Dr Hera Lee of KIT's Meteorological and Climate Research Institute in Garmisch-Partenkirchen says," Many photos published by people on the online platform show that there is a large cultural significance of the scenario. In his opinion, it has to be considered on sustainable development and landscape planning. As a result, Lee and his team have developed a method to quickly and reliably use photos in social media to draw conclusions about the socio-cultural significance of the scenarios for landscape planners.

Evaluation of photos with AI: the node size reflects the frequency of certain tags, the colors stand for different thematic clusters.

Credit: Graphics: Heera Lee, KIT

His study "Mapping Cultural Ecosystem Services Mapping 2.0 - Potential and Deficiencies from Unfinished Crowded Images" includes the area of the basal basin as an example in Saxony. "Here, the Ore attracts people who want to hike, mountain biking or winter sports," says Lee. The amount of photos posted by visitors on various platforms accordingly is high. For her study, Lee and her team collected photos posted on Flickr from 01 January 2005 to 31 December 2016. During this period, 725 different people uploaded approximately 13,000 photos of this area on the online platform.

Evaluation of photographs shows high detail that how tourists and local populations use the area of the basal basin to enjoy leisure activities or to enjoy nature. Now, scenario planners can identify "hotspot" and better understand the socio-cultural importance of some areas. On this basis, they can limit access to endangered areas or create offers and facilities to support their activities.

Evaluation of Image with the help of Machine Learning and Artificial Intelligence

In order to obtain reliable and relevant statements regarding the socio-cultural use of scenarios, Lee and his team developed a new way, which also considers the content of the photo: "We just did not want to know where the photos were And when was taken, but what was photographed primarily, this was the only way to identify centers of cultural attraction in this area. "

Researchers obtained information on photos from the tags given by users at the time of publication. They show where a photo was taken and what can be seen on this. However, the researchers faced two challenges: the quality of the tag given by the users and the duration of the evaluation. Although Flickr suggests a tag for photos and videos, they are not used by all users. Since 2500 of these photographs were studied, there was no tag in them and 590 had only one tag, the team of researchers used the cloud computing platform, through the clarifier, the image recognition algorithm based on the deep-precipitation neural network, automatic tags Decided to use. This algorithm assigned 20 tags to each picture. Altogether, 2317 individual tags were placed on approximately 13,000 photos.

read more:Manage Disk Partitions Like A Pro With These Disk Partition Software

Lee said, "We then believed that the same photos have the same tag. If the same tags are on multiple photos, then we consider it to be an indicator of related photos together." With the help of social network analysis, researchers divided the tag into nine thematic groups and assigned photos to one or more of these groups. Close scenes of flowers or butterflies were assigned to the cluster "survival", because they showed the objects present in the study areas. Lee and his team then examined the accuracy of the law through manual evaluation: About 20% of the tags were incorrectly assigned. However, researchers hope that the further development of artificial intelligence will significantly improve the reliability of the assignment. In addition, Lee and his team compared the results of the empirical study of cultural use of agricultural areas with similar results.

read more:Wireless 'pacemaker for the brain' could offer new treatment for neurological disorders

The developed method also accelerates the assessment. Approximately three hours of allotment of tags to about 13,000 photos took place. For comparison: Manually, one person is able to tag approximately 140 photos per hour. "With this method, we were able to classify and allocate photos and photos quickly and quickly from social media," says Lee. The law can also be transferred to other areas, photos of which have been published on the Internet. The European Union and Germany's Ministry of Education and Research funded the research of Hera with two projects, OPERAS, and Condit of Biodiversity 2011 respectively.