/Visual-Similarity-Based-Recommendation-System

It is a very basic recommender system that shows products similar to a given product based on their visual similarities. The pre-trained CNN models from Keras are used to extract the image features. To calculate the similarity between images cosine similarity is calculated for these image feature vectors. VGG16, ResNet50, VGG19, Xception, MobileNet models are used as feature extractors on the same corpus of images, and results are visualized separately for each model.

Primary LanguageJupyter Notebook

Visual-Similarity-Based-Recommendation-System

  • It is a very basic recommender system that shows products similar to given product based on their visual similarities.
  • Pre-trained CNN model from Keras are used to extract the image features using Transfer Learning. To calculate similarity between images cosine similarity is calculated for these image feature vectors.
  • VGG16, ResNet50, VGG19, Xception, MobileNet models are used as feature extractors on same corpus of images and results are visualised seperately for each model.

Dataset

[https://www.kaggle.com/olgabelitskaya/style-color-images]