Summary of Learning to Detect Cloud and Snow in Remote Sensing Images From Noisy Labels, by Zili Liu et al.
Learning to detect cloud and snow in remote sensing images from noisy labels
by Zili Liu, Hao Chen, Wenyuan Li, Keyan Chen, Zipeng Qi, Chenyang Liu, Zhengxia Zou, Zhenwei Shi
First submitted to arxiv on: 17 Jan 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 This research proposes a novel approach to detecting clouds and snow in remote sensing images by addressing the issue of noisy labels in existing datasets. The complexity of scenes and diversity of cloud types in these images lead to inaccurate labels, which can introduce unnecessary noise into training and testing processes. To mitigate this, the authors construct a new dataset and develop a curriculum learning paradigm that guides the model in reducing overfitting to noisy labels. They also design an evaluation method to alleviate performance assessment bias caused by noisy labels. Experiments on UNet and Segformer models validate the effectiveness of this approach. |
| Low | GrooveSquid.com (original content) | Low Difficulty Summary This paper solves a big problem with detecting clouds and snow in remote sensing images. Right now, most research focuses on making better model architectures, but that doesn’t fix the real issue: there are lots of mistakes in the labels we use to train these models. The authors come up with a new way to create datasets and train models that takes this noise into account. They test their method on two different types of models and show it works well. This is important because accurate cloud and snow detection is crucial for many applications, such as monitoring weather and climate change. |
Keywords
* Artificial intelligence * Curriculum learning * Overfitting * Unet
