In this example, we will implement one of the previously seen examples, replacing the last stages of a pretrained convolutional neural network and training the last stages for a new set of elements, applying it to classification. It has the following advantages:
- It builds upon models with proved efficiency for image classification tasks
- It reduces the training time because we can reuse coefficients with an accuracy that could take weeks of computing power to reach
The dataset classes will be two different flower types from the flower17 dataset. It is a 17-category flower dataset with 80 images for each class. The flowers chosen are some common flowers in the UK. The images have large scale, pose, and light ...
