> Further, the resolution is not very good. They can currently trace back cars and humans walking, but the features are not great. Yes, it's pretty hard to eve

> Further, the resolution is not very good. They can currently trace back cars and humans walking, but the features are not great. Yes, it's pretty hard to even tell it's a human from aerial image. This is how humans look at 7cm/pixel (DJI drone at 150m altitude): https://drive.google.com/file/d/0B9P9V4RHDRkkalJYcmZWTG1JeFk/view?usp=sharing https://drive.google.com/file/d/0B9P9V4RHDRkkalJYcmZWTG1JeFk..., https://drive.google.com/open?id=0B9P9V4RHDRkkTEN6Y0hQVTRIbGc https://drive.google.com/open?id=0B9P9V4RHDRkkTEN6Y0hQVTRIbG.... Shameless plug: We don't plan to do any surveillance, but at tensorflight.com we automatically detect a few object classes and our automatic human detector is almost ready. https://twitter.com/tensorflight/status/843606859399405568 https://twitter.com/tensorflight/status/843606859399405568 , please get in touch at kozikow@tensorflight.com if you are interested in this space.