Context
Detecting logos from brands in images can help with targeted advertising. Try out this technique with a simple image classification task to classify English Premier League Football/Soccer Club Team Logos.
If you love watching soccer and computer vision tasks this is a good beginner dataset for you!
Content
The companion to this dataset is my Top 5 Football Leagues Club Logos Dataset, where I compiled the basic logo images. For this dataset, I just created 1000 augmented images for each of the 20 team logos.
Coming Soon!!! : Once you've trained your models with the images from '.../epl-logos-big/', test out if your models can go on to detect the logos in real world mages found in the '/test/' directory. Explore why this approach works or does not work and what could be changed about the dataset itself.
Summary of variables:
- list of the teams = ['arsenal', 'aston-villa', 'brentford', 'brighton', 'burnley', 'chelsea', 'crystal-palace', 'everton', 'leeds', 'leicester-city', 'liverpool', 'manchester-city', 'manchester-united', 'newcastle', 'norwich', 'southampton', 'tottenham', 'watford', 'west-ham', 'wolves']
- dictionary of the teams and corresponding class number = {'arsenal':0, 'aston-villa':1, 'brentford':2, 'brighton':3, 'burnley':4, 'chelsea':5, 'crystal-palace':6, 'everton':7, 'leeds':8, 'leicester-city':9, 'liverpool':10, 'manchester-city':11, 'manchester-united':12, 'newcastle':13, 'norwich':14, 'southampton':15, 'tottenham':16, 'watford':17, 'west-ham':18, 'wolves':19}
Acknowledgements
I do not own the rights to the team logos, but they are publicly available and can be used in this context. It should go without saying, but do not use the logos from this dataset for your own business or commercial endeavors. Perfectly fine to use for this project though!
Inspiration
MNIST digit classifiers are overdone by now. Looking to provide simple image classification datasets here on Kaggle that use different and useful data. Also I love watching EPL - Go Man U!