- Implemented a logistic regression model for image classification of cats vs dogs.
- Published a detailed blog post explaining the bits and piecws of the entire process.
- The dataset used was the cats vs dogs image classification dataset from Kaggle.
- The dataset contains 25000 images in all and I split them as 20,000 in the training set and 5000 in the test set.
- Used Mean Squared Loss for Gradient Descent optimization and backpropagation.
- The model trained for 5000 epochs at a learning rate of 0.003 achieves 61% test set accuracy.
- Run the
preprocessing
notebook to convert the images into our actual dataset i.e. the one we will feed the model. - Go through the
logistic_regression_with_neural_networks_mindset
notebook for the training process.