Lesson1 - How train our on image set on Colab

Part 1 (2019)
1

Hi Team,

I just finished by first lesson and trying to build a image classifier. I created two folder with label names which i want to classify (Helmet & WHelmet).

  1. Now should I Untar the data using untar_data or use ImageDataBunch directly?
  2. If we have Train and Validation folder, by using ImageDataBunch from folder function will it create its own Validation Data set or use it from Validation Folder which is already presented in the folder.

Any help would be really appreciated.

Shibu

2
  1. You have to download your images before you can create a databunch with them.
  2. If you use ImageDataBunch.from_folder() you can include a valid_pct= and include a percentage between 0 and 1 to have it randomly pick images to include in the validation set if your dataset isn’t already split. What dataset are you using?