Summary of Star: a First-ever Dataset and a Large-scale Benchmark For Scene Graph Generation in Large-size Satellite Imagery, by Yansheng Li et al.
STAR: A First-Ever Dataset and A Large-Scale Benchmark for Scene Graph Generation in Large-Size Satellite Imagery
by Yansheng Li, Linlin Wang, Tingzhu Wang, Xue Yang, Junwei Luo, Qi Wang, Youming Deng, Wenbin Wang, Xian Sun, Haifeng Li, Bo Dang, Yongjun Zhang, Yi Yu, Junchi Yan
First submitted to arxiv on: 13 Jun 2024
Categories
- Main: Computer Vision and Pattern Recognition (cs.CV)
- Secondary: Artificial Intelligence (cs.AI)
GrooveSquid.com Paper Summaries
GrooveSquid.com’s goal is to make artificial intelligence research accessible by summarizing AI papers in simpler terms. Each summary below covers the same AI paper, written at different levels of difficulty. The medium difficulty and low difficulty versions are original summaries written by GrooveSquid.com, while the high difficulty version is the paper’s original abstract. Feel free to learn from the version that suits you best!
| Summary difficulty | Written by | Summary |
|---|---|---|
| High | Paper authors | High Difficulty Summary Read the original abstract here |
| Medium | GrooveSquid.com (original content) | Medium Difficulty Summary Scene graph generation (SGG) is a crucial aspect of satellite imagery (SAI), enabling the understanding of geospatial scenarios from perception to cognition. In large-size very-high-resolution (VHR) SAI, objects exhibit significant variations in scale and aspect ratio, making it essential to develop models that can holistically conduct SGG. The paper constructs a novel dataset, STAR, containing over 210K objects and 400K triplets, designed for SGG in large-size VHR SAI. To realize SGG in this domain, the authors propose a context-aware cascade cognition (CAC) framework, consisting of object detection (OBD), pair pruning, and relationship prediction. The CAC framework is specifically tailored to understand SAI regarding OBD and SGG. Additionally, the authors release an SAI-oriented SGG toolkit featuring about 30 OBD and 10 SGG methods that require further adaptation on their challenging STAR dataset. |
| Low | GrooveSquid.com (original content) | Low Difficulty Summary This paper helps us better understand satellite images by creating a new way to organize information in these images. This is important because satellite images can be very big and have many different objects, making it hard for computers to make sense of them. The authors created a special dataset with over 200,000 objects and 400,000 relationships between those objects. They also developed a new method that helps computers understand the relationships between these objects in the image. This is useful because it can help us better understand geospatial scenarios, which are important for many applications like disaster response and environmental monitoring. |
Keywords
» Artificial intelligence » Object detection » Pruning
