Description:

This dataset is specifically designed for the classification of e-commerce products based on their images, forming a critical part of an experimental study aimed at improving product categorization using computer vision techniques. Accurate categorization is essential for e-commerce platforms as it directly influences customer satisfaction, enhances user experience, and optimizes sales by ensuring that products are presented in the correct categories.

Data Collection and Sources

The dataset comprises a comprehensive collection of e-commerce product images gathered from a diverse range of sources, including prominent online marketplaces such as Amazon, Walmart, and Google, as well as additional resources obtained through web scraping. Additionally, the Amazon Berkeley Objects (ABO) project has been utilized to enhance the dataset in certain categories, though its contribution is limited to specific classes.

Download Dataset

Dataset Composition and Structure

The dataset is organized into 9 distinct classes, primarily reflecting major product categories prevalent on Amazon. These categories were chosen based on a balance between representation and practicality, ensuring sufficient diversity and relevance for training and testing computer vision models. The dataset's structure includes:

18,175 images: Resized to 224x224 pixels, suitable for use in various pretrained CNN architectures.

9 Classes: Representing major e-commerce product categories, offering a broad spectrum of items typically found on online retail platforms.

Train-Val-Check Sets: The dataset is split into training, validation, and check sets. The training and validation sets are designated for model training and hyperparameter tuning, while a smaller check set is reserved for model deployment, providing a visual evaluation of the model's performance in a real-world scenario.

Application and Relevance

E-commerce platforms face significant challenges in product categorization due to the vast number of categories, the variety of products, and the need for precise classification. This dataset addresses these challenges by offering a well-balanced collection of images across multiple categories, allowing for robust model training and evaluation.

This dataset is sourced from kaggle.