Steller sea lions in a western Aleutian Islands have declined 94 percent in a final 30 years. The involved western population, found in a North Pacific, are a concentration of charge efforts that need annual race counts. Specially lerned scientists at NOAA Fisheries Alaska Fisheries Science Center conduct these surveys regulating airplanes and unoccupied aircraft systems to collect aerial images. Having accurate race estimates enables us to improved know factors that might be contributing to miss of liberation of Stellers in this area.
Currently, it takes biologists adult to 4 months to count sea lions from a thousands of images NOAA Fisheries collects any year. Once particular depends are conducted, a tallies contingency be reconciled to endorse their reliability. The formula of these depends are time-sensitive.
In this competition, Kagglers are invited to rise algorithms that accurately count a series of sea lions in aerial photographs. Automating a annual race count will giveaway adult vicious resources permitting NOAA Fisheries to concentration on ensuring we hear a sea lion’s bark for many years to come. Plus, advancements in mechanism prophesy practical to aerial race depends might also severely advantage other involved species.
Learn some-more about investigate being finished to improved know what’s going on with a involved Steller sea lion populations by fasten scientists on a investigate vessel to a western Aleutian Islands in a video below.
- June 20, 2017 – Entry deadline. You contingency accept a foe manners before this date in sequence to compete.
- June 20, 2017 – Team Merger deadline. This is a final day participants may join or combine teams.
- June 27, 2017 – Final acquiescence deadline.
All deadlines are during 11:59 PM UTC on a analogous day unless differently noted. The foe organizers haven a right to refurbish a foe timeline if they hold it necessary.
Judging Criteria: Lowest Root Mean Square Error
The aim of a foe is to count any form of steller sea lion in any photo. See a Data add-on for some-more details.
Your acquiescence record should have a following format:
Your submissions will be evaluated by their RMSE from a human-labelled belligerent truth, averaged over a columns.
How to Enter
You will initial have to emanate an comment on Kaggle.com
You can name adult to 2 submissions to be used to calculate your final leaderboard score. If 2 submissions are not selected, they will be selected formed on your best acquiescence scores on a open leaderboard.
Your final measure will not be formed on a same accurate subset of information as a open leaderboard, though rather a opposite private information subset of your full acquiescence —your open measure is usually a severe denote of what your final measure is.
You should so select submissions that will many expected be best overall, and not indispensably on a open subset.
You might contention algorithms on the Submit Predictions page. You might have adult to 5 submissions per day.
Your acquiescence should be in CSV format. You can upload this in a zip/gz/rar/7z archive, if we prefer.
Number of Predictions
We design a resolution record to have 18636 prophecy rows. This record should have a header row. Please see representation acquiescence record on the data page.
Comment this news or article