Crowdworks Wins Mobyus' Agricultural Autonomous Driving Data Project, Precisely Recognizing Unstructured Farmland
TECHWORLD ·
✦ AI Summary
Crowdworks said on the 21st that it had won Mobyus' "agricultural autonomous driving data annotation" project.
The project aims to build AI training data so autonomous agricultural machinery can distinguish its surroundings and drive even in farmland without lane markings or signs.
The scope includes pixel-level classification of cultivated land, brush, farm roads, and dirt roads, as well as reflecting the location and distance of obstacles.
Crowdworks said on the 21st that it had won Mobyus' "agricultural autonomous driving data annotation" project. The project aims to build AI training data so autonomous agricultural machinery can distinguish its surroundings and drive even in farmland without lane markings or signs.
Crowdworks will begin building data tailored to the unstructured environments of agricultural sites. The scope includes pixel-level classification of cultivated land, brush, farm roads, and dirt roads, as well as reflecting the location and distance of obstacles.
The focus of the project is improving recognition accuracy in unstructured environments. Agricultural autonomous driving has fewer driving basis elements than general roads, and examples of missing elements include lane markings and signs.
In addition, agricultural sites have uneven boundaries between cultivated land and the surrounding environment, and they change significantly with the seasons and weather. Noise caused by dust and vibration during driving also makes data building more difficult.
Mobyus develops autonomous driving and autonomous operation technologies for agricultural machinery. The company has autonomous driving integrated control, "AutoGaia," and supplies autonomous driving and integrated control technologies based on its in-vehicle infotainment (IVI) systems business.
Crowdworks is establishing standards for producing autonomous driving training data suited to the shape and characteristics of farmland. To this end, Crowdworks is designing data processing guidelines based on the physical characteristics of farmland.
Crowdworks plans to support the autonomous driving system in identifying drivable areas and surrounding environments in unstructured farmland. It also plans to further subdivide objects and regions in video data, applying Semantic Segmentation as a core technology for the work.
Semantic Segmentation performs pixel-level classification of images for cultivated land, brush, farm roads, and dirt roads, and turns the boundaries of each area into training data. It also includes size, location, and distance information as factors reflected in obstacle data, building training data so autonomous agricultural machinery can make judgments such as slowing down or making an emergency stop after detecting an obstacle.
These tasks are being pursued based on Crowdworks' experience applying 2D bounding box, 3D LiDAR cuboid, and OCR data processing technologies in the autonomous driving sector. Crowdworks also has experience in collecting edge-environment data and in building autonomous driving solutions for maritime vessels.
Crowdworks said it plans to use this project as an opportunity to expand the application of its autonomous driving data-building experience from automobiles and mobility to agriculture. To that end, it aims to secure the data-building capabilities needed for actual agricultural machinery driving in farmland and plans to expand the range of data needed for surrounding-environment recognition, obstacle recognition, and route judgment. The company said this trend also sets the direction of business expansion toward the agricultural physical AI sector.
Crowdworks said farmland is a representative unstructured environment without signs. It also said work on farmland requires more advanced annotation standards than general autonomous driving. Crowdworks said it aims to contribute to the development of physical AI technology in agriculture based on its data processing technologies.
Source: TECHWORLD · Kim Seung-gi
Original: https://www.epnc.co.kr/news/articleView.html?idxno=407186
References
This article was produced with the help of an automated content generation algorithm.
Source: TECHWORLD
View originalThis article was summarized and organized by BizCrush based on the original article from TECHWORLD. For exact quotations and full details, please refer to the original article.