Skip to content

Latest commit

 

History

History
15 lines (8 loc) · 1004 Bytes

README.md

File metadata and controls

15 lines (8 loc) · 1004 Bytes

Sneaker Classifier CNN

See the CNN in action: sneakername

A convolutional neural network that classifies up to 900 sneakers with 90% accuracy. Trained using the fast.ai libary.

How this was build

First I scraped the StockX website to get a large csv file of product data. I modifed an existing script from zpencerguy to create a new script that worked on the StockX website. See here: StockX-scraper.

Next, I cleaned the data and put all the product names in a text file, and wrote an Python program to download 100 images per sneaker from Google. This got me 90k images total. Then, I cleaned the data manually which left me with 52k usable images.

Finally, I used the Fastai libary to train a convolutional neural network on the images and got an decent accuracy of 90%.

It's far from perfect, but it was a fun experiment.