Shulou(Shulou.com)11/24 Report--

Led by Institute of Aerospace Information Innovation, Chinese Academy of Sciences (CAS), United China Science and Technology Star Map Co., Ltd. (China Science Star Map), China Science Star Map Digital Earth Hefei Co., Ltd. (Star Map Earth), IEEE GRSS and IEEE GRSS IADF jointly organized the AI-empowered Flood Map Marathon (AI Powered Flood Mapathon) on October 16, 2023.

The star map earth brain engine product developed by Mapathon based on the star map of China Science and Technology is a multi-person cooperative activity to create a map. Participants can develop and use some new tools or methods together to improve the accuracy and practical value of mapping.

AI Powered Flood Mapathon contest

This competition is the first flood map marathon in China, which draws important feature elements of flood-prone areas by collaborative relay according to satellite images, in order to call attention to natural disasters such as floods all over the world, and contribute to disaster early warning and humanitarian assistance by participating in the mapping of flood-prone areas.

Significance of the competition

Flood is the most common and serious natural disaster in the world. According to the disaster Human loss report released by the United Nations disaster Prevention and reduction Program (UNDRR), between 2000 and 2019, a total of 3254 floods were recorded worldwide, accounting for 44% of the total number of disaster events, resulting in 1.65 billion people affected, 104614 deaths and economic losses of US $651 billion, more than any other type of disaster. (data from: United Nations Office for Disaster Risk Reduction)

In addition, the frequency of future floods is likely to increase further as a result of global climate change and changes in urbanization. Mapping flood-prone areas is essential for flood early warning and disaster relief assistance. On the one hand, providing valuable data for the training of flood early warning model is helpful to study and analyze the triggering factors of past flood events and to warn future events; on the other hand, the high-precision geospatial information provided can help rescue workers to quickly assess the damage degree of disaster areas, make reasonable rescue plans, and win valuable time for rescue and disaster relief.

Main task

The main task of the competition is to draw important features in flood-prone areas according to satellite images. the system supports manual rendering, automatic rendering and combination of artificial intelligence technology. the purpose of this paper is to use artificial intelligence technology to promote the development of map labeling to a more refined and efficient direction, and to help the further development of flood disaster early warning and rescue.

Characteristics of the competition

-AI-enabled mapping-

The AI enablement of the Star Map Earth brain engine is a special feature of this competition, which runs through the task, process, evaluation and so on. On the one hand, the contestants can obtain the results of preliminary automatic rendering of the selected area by using the intelligent tagging tools in the tagging system, so as to draw further based on this result, and with the improvement of the quantity and quality of labeling, the mapping ability of this intelligent tagging algorithm is expected to develop continuously; on the other hand, the map can be intelligently segmented with the help of the function of SAM to help improve the labeling efficiency. In addition, you can download the satellite image block to be labeled, run your own intelligent algorithm to complete the tagging, and then upload it to the entry system.

-excellent visualization-

In order to make it more interesting, if the contestants draw the same satellite image block many times, the system can generate GIF animation according to the tagging results submitted each time, which can visually show the tagging change process of the image block, and facilitate the contestants to share freely on the mainstream social media platform.

-Open and free data resources-

Contestants are free to download satellite images and other data for academic research, or they are free to share their drawings with others.

Task design

The competition provides participants with high-resolution RGB satellite images at two time points before and during the flood, which are used to draw important features in flood-prone areas. The main feature elements include background, buildings, roads, water, wasteland, forest and farmland.

Map (main feature elements)

In order to meet the different needs of the contestants, the system also provides three drawing methods for choice: the first is to draw manually from beginning to end; the second is to draw automatically using advanced artificial intelligence algorithms; and the third is a combination of the two.

Process design

The competition advocates contestants to efficiently complete the mapping of flood-prone areas through cooperation and relay, which mainly includes the following steps:

-step 1-

The contestants register and log in to the online visual tagging system, click on any unmarked or partially annotated satellite image and get the task and enter the tagging system.

-step 2-

In the relay race, the contestants who get the first bar of block labeling can choose to label manually, based on the default artificial intelligence algorithm or the automatic drawing results of their own algorithms, while non-first-leg contestants can only continue to label manually according to the previous marking results of the block, or with the aid of SAM-assisted segmentation function.

-step 3-

After the completion of the marking, the contestants release the block by submitting the task, and then go to pick up the marking tasks of other blocks, so that other contestants can continue to label the block.

Evaluation design

The organizing committee uses the mIoU method to evaluate the marking work of the contestants. Because relay cooperation is very encouraged in this competition, after the contestants mark each block, the system will calculate a corresponding score according to the improvement of the quality of the block labeling. finally, it is ranked according to the cumulative score of all the marked blocks of the contestants.

At the same time, the competition will provide contestants with rich prizes:

Top 15: 130mm large aperture professional stargazing telescope

16-50: circular induction light flow sand painting AI intelligent Bluetooth audio

51-100th: luminous DIY three-dimensional building block globe

101-200: globe penholder

In addition, each winner will receive a certificate issued by IEEE GRSS in recognition of their great contribution to mapping in this competition.

Picture: prize

Specific information of the competition

The AI Powered Flood Mapathon competition officially started on October 16, 2023, the system was closed on January 15, 2024, and the winners were announced on January 30, 2024.

Welcome to subscribe "Shulou Technology Information " to get latest news, interesting things and hot topics in the IT industry, and controls the hottest and latest Internet news, technology news and IT industry trends.