Sourcing Images, Build Model, and Deployment
As an exercise for Chapter 2: Production of Fast.ai, we are to outsource our images and build our own image classifier. Outsourcing from the internet haven’t been the best experience. This time, one decides to build an eagle classifier. However, just like any other images, “eagle” can have different meaning.
For example, searching for “sea eagles” haven’t been much luck as half of the image is as expected, “real” sea eagles, while the other half represents the football team Manly Warringah Sea Eagles.
And searching for snake eagles, well you get this image.
Changing their names to “fish eagles” and “serpent eagles” gives better results.
The full project can be found here, along with the binder link:
Upload an image and play with it. Have fun!