This is not a computer vision problem, because spatial correlation between geographic cells is destroyed. Figure modified from Mignan & Broccardo (2019c), representing the deep learning workflow of DeVries et al. – Instead of having nx times ny features for n mainshock samples, the authors used roughly nx times ny times n geographic cell samples (and 12 stress-based features).
Latest posts by Alka Tripathy-Lang, PhD (see all)
- Magnitude-6.5 earthquake rattles Nevada and California - May 15, 2020
- Volcanic lightning helps aviators avoid hazardous ash - May 1, 2020
- Slow moving fault pieces may limit large Himalayan quakes - March 23, 2020