fastai-api-notes

This is an attempt to condense the FastAI package info. Think of it as a cheatsheet or notes for experienced FastAI developers to directly refer to the code needed for their problem

Current Set of Imports

from fastai.vision.all import *
from fastai.text.all import *

# we will not cover these two. scikit-learn / other standard python packages can be used for such use cases
from fastai.collab import *
from fastai.tabular.all import *

As mentioned in the docs, fastai’s applications all use the same basic steps and code:

• Create appropriate DataLoaders
• Create a Learner
• Call a fit method
• Make predictions or view results.

Know your Deep Learning Problem: Problems covered by FastAI API

Vision:

• Classification
  • Single-label
  • Multi-label
• Image Localization
  • Segmentation (predict the class of each pixel of an image)
  • Coordinate Predictions (predict one or several key points on an image)
  • Object Detection (draw a box around objects to detect)