2019/08/12 by Venkatesh Umaashankar, Umaashankar, Venkatesh, Girish Shanmugam S +3 · 1 citation
Computer Science · Engineering · Mathematics · #Advanced Image and Video Retrieval Techniques #Artificial intelligence #Benchmark (surveying) #Cartography #Categorization #Clothing #Computer science #Data mining #Data science #Generative Adversarial Networks and Image Synthesis #Geography #Image Retrieval and Classification Techniques #Industrial Vision Systems and Defect Detection #Information retrieval #Machine learning #Mathematics #Natural language processing #Product (mathematics) #Statistics #Taxonomy (biology) #Unavailability #Video Analysis and Summarization #cs.CV #cs.LG #stat.ML
paper · pdf · doi:10.48550/arxiv.1908.08984
published in arXiv (Cornell University) (Cornell University) · preprint
arxiv created 2019/08/12 · openalex publication_date 2019/08/12 · arxiv updated 2019/08/27 · openalex created_date 2022/07/28 · openalex updated_date 2026/08/05
In E-commerce, it is a common practice to organize the product catalog using\nproduct taxonomy. This enables the buyer to easily locate the item they are\nlooking for and also to explore various items available under a category.\nProduct taxonomy is a tree structure with 3 or more levels of depth and several\nleaf nodes. Product categorization is a large scale classification task that\nassigns a category path to a particular product. Research in this area is\nrestricted by the unavailability of good real-world datasets and the variations\nin taxonomy due to the absence of a standard across the different e-commerce\nstores. In this paper, we introduce a high-quality product taxonomy dataset\nfocusing on clothing products which contain 186,150 images under clothing\ncategory with 3 levels and 52 leaf nodes in the taxonomy. We explain the\nmethodology used to collect and label this dataset. Further, we establish the\nbenchmark by comparing image classification and Attention based Sequence models\nfor predicting the category path. Our benchmark model reaches a micro f-score\nof 0.92 on the test set. The dataset, code and pre-trained models are publicly\navailable at urlhttps://github.com/vumaasha/atlas. We invite the community\nto improve upon these baselines.\n