Project analysis – Rakuten France Multimodal Product Data Classification

---

Contents

• Datasets
  • Files
  • Directories
• Dataset analysis
  • X_train_update.csv
  • X_test_update.csv
  • Y_train_CVw08PX.csv
• Images

---

Datasets

Files

• X_train_update.csv: Training input
• Y_train_CVw08PX.csv: Training output
• X_test_update.csv: Test input

Directories

• image_train: Training input images
• image_test: Test input images

WARNING

All training sets must be split into training and validation sets as there are no labels provided for the test sets. They are for final validation exclusively!

Dataset analysis

X_train_update.csv

Shape: 84916, 5

Variable Name	Description	Dtype	Total Values	Missing Values	Unique Values
Unnamed: 0	Consecutive indices	int64	84916	0	84916
designation	Short product designation	object	84916	0	82265
description	longform product description with html formatting	object	55116	29800	47506
productid	unique product ID	int64	84916	0	8491
imageid	unique image ID	int64	84916	0	84916

---

X_test_update.csv

Shape: 13812, 5

Variable Name	Description	Dtype	Total Values	Missing Values	Unique Values
Unnamed: 0	Consecutive indices	int64	13812	0	13812
designation	Short product designation	object	13812	0	13812
description	longform product description with html formatting	object	8926	4886	8346
productid	unique product ID	int64	13812	0	13812
imageid	unique image ID	int64	13812	0	13812

---

Y_train_CVw08PX.csv

Shape: (84916, 2)

Variable Name	Description	Dtype	Total Values	Missing Values	Unique Values
Unnamed: 0	Consecutive indices	int64	84916	0	84916
prdtypecode	Product type code	int64	84916	0	27

---

There are 27 product classes (target variable 'prdtypecode').

Images

Naming convention: image_<imageid>_product_<productid>.jpg

Example: image_67284_product_365202.jpg

Size: 500px x 500px

Format: .jpg

image_train contains 84916 files, none missing

image_test contains 13812 files, none missing