Or try an example cover:
Judge
Based on the cover image, predict the popularity (high, medium, or low) of your book
Methodology
I created this website as my Capstone for the Machine Learning Guild Apprentice Program.
The book-cover-judger is a CNN Classifier built via ResNet34 transfer learning with the fastai library. The model was trained on book cover images and review data scraped from Goodreads, predicting popularity tertiles (High, Medium, or Low) based on ratings count.
The prediction accuracy rate is 44%. While admittedly low, it's still better than random! (33% for classifying into three segments)
View the training notebooks and data pipeline on GitHub.
About Me
Analyst looking for data science staffing opportunities!
Grant Tesdahl
Analyst // MinneapolisI'm actively seeking roles that extend my knowledge of data science to the development of robust ML models and the infrastructure to deploy them.
When I'm not reading books or judging their covers, I enjoy skiing, listening to Freakonomics Radio podcasts, and rock climbing.