Tip #2: Do Transfer Learning When You Can

In the context of DNN modeling of image data, "transfer learning" refers to the process of using a pre-trained model (such as VGG, ResNet, or Inception) and fine-tuning it with your limited data. The idea is that image data have similar properties, and the first few layers of CNN-based models are designed to capture general features that should work well for all images, regardless of the domain.

This approach often works very well in practice. In fact, it is safe to say that for more than 95% of biomedical image applications (in radiology and pathology), deep learning models would not be successful without the models developed in image classification tasks like identifying dogs and cats.

Take advantage of pre-trained models whenever you can. They can save your day.

Read next tip: Watch Your Own Data


Need assistance in your AI/deep learning project? We may be able to help. Take a look at the intro to our bioinformatician team, see some of the advantages of using our team's help here, and check out our FAQ page!

Send us an inquiry, chat with us online (during our business hours 9-5 Mon-Fri U.S. Central Time), or reach us in other ways!