🍕Products-10K🍕 — SOTA Dataset for Products Image Recognition
Products-10K is a large-scale dataset for product recognition. The dataset consists of images of consumer goods. The data was collected to improve the quality of models for recognizing goods on warehouse shelves.
Why is it important?
Product recognition is an urgent problem for Computer Vision due to the wide variety of products and their visual similarity, that currently lacking in high-quality and diverse datasets for training. The existing datasets with product images are either too small or contain quite a noisy markup.
However, recently there was published an article at Arxiv, called “Products-10K: A Large-scale Product Recognition Dataset” that turns to be a new state-of-the-art dataset for solving an object recognition issue for product images. Products-10K consists of 10,000 product categories that are frequently purchased by user at chinese shop JD.com
Dataset desctiption
Images in Product-10K were collected from online store photos and user images. Users attached images to comments on a purchased product.