The purpose of this project is to benchmark text and image processing techniques to build features as input for a classifier that should help to sort products into their categories.
Bag of words, Bag of visual Words.
We use SIFT (equivalent to SURF, ORB) and CNN VGG16 for image features extraction (after assessment of CNN for rough prediction).
We use nltk, tf-idf, and gensim for text features extraction.
We can combine many techniques and upgrade to multi-label predictions.
EtienneLardeur/P6_ClassificationEngine
Enhance product classification into market-place categories, through text and image processing.
Jupyter Notebook