Computer vision has improved rapidly over the past decade due to a few intersecting trends:
- Research advancements in machine learning
- Petabytes of image data freely available online
- Open Source culture within the AI community
One technique available to data scientists is transfer learning. This is particularly useful for image recognition, models for which typically require a large amount data and extensive computation to train.
Instead of training a model from scratch, we can import the weights from another model as a basis for our specific use case.
InceptionV3 is one of the most advanced models for computer vision currently available. Created by Google Research members Szegedy et. al. The graphic above shows the complex architechture of this very deep network.
A link to the paper can be found here: https://arxiv.org/pdf/1512.00567.pdf