Awesome

Dimension Extraction Dataset

This dataset contains the golden finetuning and test dataset (BBC_GDS) used in the paper D-Extract: Extracting Dimensional Attributes from Product Images, published at WACV 2023. The dataset (final_dim_extract_data.csv) contains 69 product types (PTs) with approximately 150 training images and 50 test images for each PT. For each image datapoint, we provide all available manually annotated values in the image in both metric and imperial units (if available in the image). The relevant attributes for each product type are described in the PT_metadata.csv file. Each datapoint is provided with its image link instead of the original link due to copyright.

Figure 1. Example of a product image with dimensional attributes. According to the established definition (Section 2), the table’s length, width and height are 23.5 inch, 68 inch and 36.5 inch respectively.

If using our dataset, please cite the following:

@Inproceedings{GHOSH2023, author = {Pushpendu Ghosh and Nancy Wang and Promod Yenigalla}, title = {D-Extract: Extracting dimensional attributes from product images}, year = {2023}, url = { https://www.amazon.science/publications/d-extract-extracting-dimensional-attributes-from-product-images }, booktitle = {WACV 2023}, }

Security

See CONTRIBUTING for more information.

License

This project is licensed under the Apache-2.0 License.