FastAI

Simple Solutions for Complex Challenges

FastAI is a user-friendly deep learning library that simplifies complex machine learning tasks. It provides pre-built models and tools to make it easier for beginners to create powerful AI applications without needing extensive coding knowledge.

Higlights:

In general it is a good library to be used with PyTorch in order to make fast ready to use Deep Learning models. I created a model capable of identify if an image was or was not Pizza. below i will write what were the advantages and disadvantages that i found so far:

Advantages

  • Very fast and easy way of create DNNs

  • We can select pre trained architectures or train our own

  • We have some good built in functions to prepare data and visualize results

  • It seems that it is a widely used library and has a good quantity of possible different uses

  • Not so bad documentation (not extraordinary either)

Disadvantages

  • It looks too short for certain more difficult tasks

  • it could have more pre-trained architectures directly implemented

  • Do not know if it is that usefull to use if you already master PyTorch or Tensorflow

  • Sometimes if you wanna understand what is happening you need to go read the source code


Work done to test this

https://www.kaggle.com/code/franciscomesquita/pizzaornotpizza-pytorch-fastai-97-acc-effnet


Score: 7