- First performed transfer learning and extract features from Covid-19 chest X Rays dataset.Used the Keras deep learning library and the ResNet50 network (pre-trained on ImageNet). Used ResNet50 to forward propagate images to a pre-specified layer. This output activations of that layer and treat them as a feature vector.
- The output volume of ResNet50 is 7 x 7 x 2048 = 100,352-dim. Assuming 32-bit floats for our 100,352-dim feature vectors, that implies that trying to store the entire dataset in memory at once would require 10.03GB of RAM.
- Used the Creme library for Incremental Learning we trained a multi-class Logistic Regression classifier one sample at a time, obtained 97.6% accuracy on the Dogsvs.Cats dataset.
- The model trained on all 25,000 samples, we reach 97.6% accuracy which is quite respectable.
SrikanthIITB/Incremental-Learning-with-Keras-and-Creme
The model trained on all 25,000 samples, we reach 97.6% accuracy.
Jupyter Notebook