This page organizes papers related to artificial intelligence published around the world. This page is summarized using Google Gemini and is operated on a non-profit basis. The copyright of the paper belongs to the author and the relevant institution. When sharing, simply cite the source.
Maggie Chen, Hala Lambdouar, Luca Marini, Laura Martinez-Ferrer, Chris Bridges, Giacomo Acciarini
Outline
Methane is a potent greenhouse gas and a major contributor to climate change, making timely detection crucial for effective mitigation. Satellite-based machine learning (ML) can enable rapid detection while reducing downlink costs, supporting faster response systems. Existing methane detection methods often rely on image processing techniques such as orthophotometry to correct geometric distortions and matched filters to enhance plume signals. In this paper, we present a novel approach that bypasses these preprocessing steps by using unorthophotometric data (UnorthoDOS). We find that ML models trained on this dataset achieve comparable performance to models trained on orthophotometric data. Furthermore, we demonstrate that models trained on the orthophotometric dataset outperform the matched filter baseline (mag1c). We are releasing two ML-ready datasets ( https://huggingface.co/datasets/SpaceML/UnorthoDOS) and code ( https://github.com/spaceml-org/plume-hunter) consisting of model checkpoints and ortho-projected and non-ortho-projected hyperspectral images obtained from the Earth Surface Mineral Dust Source Investigation (EMIT) sensor .
Takeaways, Limitations
•
Takeaways:
◦
We present a novel approach to training ML models using unprocessed data, potentially simplifying preprocessing steps and reducing computational costs.
◦
We demonstrate that ML models trained on orthogonalized data outperform conventional matching filter-based methods.
◦
We have made publicly available ML model checkpoints, orthogonalized and non-orthogonalized datasets, and code for methane detection research to facilitate reproducibility and scalability of the research.
•
Limitations:
◦
There is no specific mention of Limitations in the paper.
