Skip to content

Commit

Permalink
Browse files Browse the repository at this point in the history
Update readme and add imagenet labels
  • Loading branch information
Kevin Scott committed Aug 8, 2018
1 parent 04802ac commit 6b5202d
Show file tree
Hide file tree
Showing 3 changed files with 1,012 additions and 2 deletions.
12 changes: 11 additions & 1 deletion README.md
Expand Up @@ -12,7 +12,7 @@ It's important that the training images you use are as similar as possible to th

For instance, if you're training an image classifier to recognize pictures of fruit and your training data consists of well let brightly colored fruit but your users upload grainy, low-light shots, your accuracy will suffer.

## How to Build Your Own Dataset
## Organizing your Dataset

To put together a dataset for training an image classifier, you need to organize your data like so:

Expand All @@ -26,6 +26,16 @@ To put together a dataset for training an image classifier, you need to organize
- /class-two
```

## Populating your dataset

Shoot for 80% training data to 20% validation data.

Some reference links:

* https://stackoverflow.com/questions/13610074/is-there-a-rule-of-thumb-for-how-to-divide-a-dataset-into-training-and-validatio

* https://www.researchgate.net/post/Is_there_an_ideal_ratio_between_a_training_set_and_validation_set_Which_trade-off_would_you_suggest

## Example
Let's say you want to recognize cats vs. dogs. Your folder structure would look like:

Expand Down

0 comments on commit 6b5202d

Please sign in to comment.