Pistachio Classification

In this implementation, we followed three main approaches:

1- transer learning: choosing a powerful pre-trained model (inceptionV3) to extract features from images using PyTorch

2- extracting color based and texture-based features from the dataset and concating them with already obtained shape and morphological features :
- 1 - color based ---> average RGB method ---> features added : (1- Red avg 2- Blue avg 3- Green avg)
- 2 - texture based ---> Gray level co-occurrence matrices(GLCM)---> features added : (1- contrast 2- dissimilarity 3- homogeneity 4- energy 5- correlation)

3- hybrid method of 1 and 2 random forest was used as the main classifier.