Recognizing and Localizing Right Whales with Extremely Deep Neural Networks

Very interesting write up. I don't know a lot about machine learning, but these kinds of articles are a great starting point to learn about different concepts.

I did find the whale localizer step to be a bit strange, using more machine learning and labeled datasets instead of using unsupervised image processing techniques. Seems like a black and white and oblong whale should be simple to extract and orient without needing specialized data sets. It seems like I'm not the only one to think this since the author remarks about the number of these different kinds of localizers in the forum.

Wonderful write-up. Posts like these give just the right amount of detail and high-level insight.

I especially appreciated the linear flow of the approaches section. The successes and roadblocks presented provide useful context into why certain decisions were made, which is too often missing from many academic papers.

This is a really fantastic deep dive into the practicalities of solving one of these Kaggle competitions. I'm at a personal level of graduating from taking online classes to trying to apply ML techniques to real problems, and found this really helpful.

Also, write-up from the winner of the competition: http://deepsense.io/deep-learning-right-whale-recognition-ka...

They use a very similar technique, I like the emphasis on the white pattern on the head to help distinguish each whale.

I wanted to do something like this for years. Is a classic problem in cetology.