The main goal of this example is to demonstrate the use of the MATLAB functionality for scene classification solution using a subset of the MIT Places dataset and a pretrained model, Places365GoogLeNet.
The code is structured in four parts:
- In "Part 1", we build a simple CNN from scratch, train it, and evaluate it.
- In "Part 2", we use a pretrained model, Places365GoogLeNet, "as is".
- In "Part 3", we follow a transfer learning approach that demonstrates some of the latest features and best practices for image classification using transfer learning in MATLAB.
- Finally, in "Part 4", we employ image data augmentation techniques to see whether they lead to improved results.
This example should be easy to modify and expand to the user's needs.
- The dataset used in this example (a subset of the Places365-Standard dataset) can be found at: https://www.dropbox.com/s/addp3xkw1g0ua7v/MITPlaces.zip?dl=0
- Companion blog post appearing soon at: https://blogs.mathworks.com/deep-learning/
Oge Marques (2020). Scene Classification Using Deep Learning (https://www.mathworks.com/matlabcentral/fileexchange/73333-scene-classification-using-deep-learning), MATLAB Central File Exchange. Retrieved .