3 ms·
Shouldn't this be quite easy to do with existing computer vision technology though (I'd assume this is a computer vision 101 class project)? Especially consider
by curiousDog 13y ago
Shouldn't this be quite easy to do with existing computer vision technology though (I'd assume this is a computer vision 101 class project)? Especially considering this is the ocean. The images were taken march 16nth, so they probably were running them through their software.
- roel_v 13y agoDepends on what you call 'quite easy'. In the terminology of http://fishbowl.pastiche.org/2007/07/17/understanding_engineers_feasibility/ http://fishbowl.pastiche.org/2007/07/17/understanding_engine... , it's still a 'hard' problem. There are tens of thousands if not an order of magnitude more objects and reflections of that size in the ocean. It certainly is not a 'computer vision 101' project. 'Computer vision 101' is 'do edge detection using opencv', not 'build a complete system that interfaces near-realtime with satellite imagery sources that can do very detailed analysis, but without a real definition of what we're looking for, on huge amounts of data'.
- Houshalter 13y agoPossibly. We don't really have a dataset of "images of crashed planes" for it to train on though. Perhaps you could use unsupervised learning to find images and image segments that are anomalous. However I expect you'd mostly get pictures of boats, planes, small islands, weird waves, floating garbage, etc. But that does significantly narrow down your search at least.