Project 6: Automated classification of goods from text and images

Keywords: unsupervised learning, NLP, computer vision, deep learning, transfer learning

Goal

This project aims at categorizing products with unsupervised learning. Data include names and descriptions of products as well as photos. There are 1050 items to classify in 7 categories.

What you need

The csv dataset directly on the github repo (flipkart_com-ecommerce_sample_1050.csv).

Main librairies used:

• Python 3.8.10
• Matplotlib 3.4.2
• Pandas 1.1.3
• Nympy 1.19.2
• Seaborn 0.11.0
• Gensim 3.8.3
• Tensorflow 2.7.0
• Plotly 5.4.0

FILES

• P6_preprocess_clustering.ipynb: main jupyter notebook file
• fonctions_eda.py: basic exploratory data analysis functions
• flipkart_com-ecommerce_sample_1050.csv: text data and link to images
• images directory

Approaches

• NLP: bag-of-words, tf-idf, word2vec, doc2vec
• Computer vision: ORB, VGG16