search

microsoft / arcticseals

A deep learning project in cooperation with the NOAA Marine Mammal Lab to detect & classify arctic seals in aerial imagery to understand how they’re adapting to a changing world.
MIT License
33 stars 14 forks source link

Validation (and API for using our tools) #7

Open Marcel-Simon opened 6 years ago

Marcel-Simon commented 6 years ago

It might be very helpful, if we had a way to provide our model in such a way, that it is very easy to use even with little to no technical knowledge. I was thinking, that we could provide it using an API and some GUI tool, which allows using this API. What do you think?

Athapan commented 6 years ago

What kind of API are you think?

Marcel-Simon commented 6 years ago

I was thinking about the latter, i.e. somebody can upload (a set of) images and the API returns locations of seals. The same goes for the classification model, i.e. somebody can upload a (color) image of an hotspot detection, and the API decides whether this is a seal or not.

pbaer commented 6 years ago

Some thoughts on how we could structure this that makes it easy for us to evaluate various approaches, and potentially helps package things up for NOAA in a somewhat consumable way.

Input: a folder of thermal and color files (using the filename pattern used by NOAA to indicate timestamp, color vs. thermal, etc.).

Output: a "hotspots.csv" file of hotspot detections (similarly using the same CSV schema NOAA gave us), written into the same folder

A nice independent task for someone to work on Tuesday would be do manage the validation set and confirm detection accuracy. I think it would be especially good for someone not directly working on a detection model to do this, to maintain objectivity :)

(updated issue title)