June 2018
Intermediate to advanced
436 pages
10h 33m
English
In this project, we saw how to solve a multi-label image classification problem. We used real Yelp images. Then we trained a CNN to predict the classes for each tagged image. In this project, the most challenging part was feature engineering, as we had to deal with not only images but also different tags and metadata. Unfortunately, we could not achieve very high accuracy.
The takeaway would be that similar approaches can be applied to solve other image datasets having multi-labels. Yet, a multiclass classification problem can be solved with minimal effort as well. All you need is to prepare the dataset such that a CNN-based model can consume it. Apart from this outlook, the project ...