DataFramed #107 The Deep Learning Revolution in Space Science
Oct 3, 2022
Justin Fletcher, an AI expert at the US Space Force, discusses the groundbreaking applications of deep learning in space science. He reveals how the Space Force monitors satellites and tracks dangerous space debris using advanced AI techniques. The conversation delves into vital skills for aspiring defense practitioners and the importance of collaboration among data scientists and domain experts. Justin also emphasizes transparency in military research and the need for effective communication of complex concepts in both technical and non-technical circles.
AI Snips
Chapters
Transcript
Episode notes
Cross-Disciplinary Team
- Justin Fletcher's team in Maui includes astronomers, computer vision specialists, data scientists, and software engineers.
- This diverse team fosters cross-pollination of ideas, leading to computer scientists learning astronomy and astronomers learning data science.
Domain Knowledge
- Applying data science techniques to new domains requires essential domain knowledge. Directly applying a ResNet to FITS images for classification won't work without data transformations.
Deep Learning Applications
- The Space Force uses deep learning for various tasks like object classification and space object detection.
- Their models range from traditional CNNs to more advanced architectures like Deformable DETR.