Tag
This paper investigates augmentation techniques for CNN-based classification of multispectral images from visible and thermal infrared cameras, using the ThermalWorld dataset to understand how different augmentations affect classification accuracy.
The article argues that AI's potential to augment human productivity is limited by humans lacking a serious context of use and being bottlenecked by internal factors where external tools can't help.