I’m teaching deep learning this semester and we are using Fast.ai’s new V2 library. It’s week 3 of the course and the students have to make an image classifier. As an example, I made a “five animals” classifier. The “five animals” are the five traditional heard animals in Mongolia, and are considered a strategic asset by the state. The animals are sheep, goats, camels, horses, and cattle.
The classifier is only about 91% accurate from testing, but it’s a great fast example of how you can get going really fast with Fast.ai and then deploy really fast with Anvil. Here is the app in action using an Uplink to perform inference:
You can get the Colab notebook here:
https://colab.research.google.com/drive/1o477I9Ojdpm2tftsZRasp5ocjqFWpZjX?usp=sharing
And here is the Anvil app:
https://anvil.works/build#clone:BXIBL2GTKQFZ3RHV=QSLPD6PFAK7GXKWK2KLNGZGL
PS: I didn’t add much markdown to the notebook as I have a separate lecture that explains all the code. However, the current Fast.ai course (course.fast.ai) goes over all of the material here if you are interested.
PPS: It’s pronounced yeaahmmdd (YAMD).