AttireClassifier

Attire Classifier is a Cultural Clothes Classifier aimed at creating a CNN Model that can classify clothes from different regions of Pakistan.

Dataset

Following Dataset was assembled and used: https://www.kaggle.com/datasets/khizertariq/pakistani-cultural-clothes

Model Architecture:

For classification a CNN Model was used that predicts outputs for four classes.

Data Cleaning:

The converter.ipynb can be used to convert files to jpg format while the datacleaningscript.ipynb can be used to clean the directories while keeping images of only the allowed formats.

GradCam:

The Code also uses Gradient-weighted Class Activation Mapping (Grad-CAM) to highlight the regions influencing the decisions of the CNN model thus taking a step towards Explainable AI.