The notebook "Squares" contains code for the Squares dataset.
The dataset contains 3 types of squares, which are visually varying in intensity. The objective of the problem is to classify the 3 types of squares using any method.
The data is not being provided but the code can be used as opensource.
The code contains:
- Histogram plot of images per channel
- CNN to classify
- Quantitative analysis of the dataset
- Feature extraction and PCA using pretrained models(VGG19,VGG16,Resnet)
- Error analysis to find patterns in the misclassified data
Part 1 contains all expereiments which can be run with the dataset
Part 2 contains sample runs which were used for analysis