README.md

metadata

language: en
tags:
  - image-retrieval
  - copydays

Dataset Card for Copydays

Dataset Description

Copydays is a dataset designed for evaluating copy detection and near-duplicate image retrieval algorithms. It contains images with various modifications to test the robustness of retrieval systems.

• copydays_original: Original, unmodified images.
• copydays_strong: Images with strong modifications (e.g., cropping, rotation, compression).

These datasets are widely used for benchmarking image retrieval systems under challenging conditions.

Dataset Features

Each example contains:

• image (Image): An image file (JPEG or PNG).
• filename (string): The original filename of the image (e.g., 200000.jpg).
• split_type (string): The type of split the image belongs to (original or strong).
• block_id (int32): The first 4 digits of the filename, representing the block ID (e.g., 2000 for 200000.jpg).
• query_id (int32): The query ID for query images (-1 for database images). Digits 5 and 6 of an image name (e.g., 01 for 200001.jpg).

Dataset Splits

• queries: Query images with modifications for evaluation. Also includes the original images.
• database: Original images used as the database for retrieval.

To tell if something is an original image or a strongly modified image, refer to a given images split_type field. An example is shown in the Example Usage below.

Dataset Versions

• Version 1.0.0

Example Usage

Use the Hugging Face datasets library to load one of the configs:

import datasets

# Name of the dataset
dataset_name = "randall-lab/INRIA-CopyDays"

# Load query images
query_dataset = datasets.load_dataset(
    dataset_name,
    split="queries",
    trust_remote_code=True,
)

# Load database images
db_dataset = datasets.load_dataset(
    dataset_name,
    split="database",
    trust_remote_code=True,
)

# Print the length of the query dataset -- should be 386, since it includes all 229 strong AND all 157 original queries
print(f"Number of query images: {len(query_dataset)}")

# You can tell if it is a strong or an original query by checking the `split_type` field on a given image
example_query = query_dataset[0]  # Get any desired query image
print(f"Example Query - Filename: {example_query['filename']}")
print(f"Example Query - Split Type: {example_query['split_type']}")

# Print the length of the database dataset -- should be 157, since it includes all 157 original images
print(f"Number of database images: {len(db_dataset)}")

Dataset Structure

• The datasets consist of images downloaded and extracted from official URLs hosted by the Copydays project.
• The copydays_original dataset contains unmodified images.
• The copydays_strong dataset contains images with strong modifications.

Dataset Citation

If you use this dataset, please cite the original paper:

@inproceedings{jegou2008hamming,
  title={Hamming embedding and weak geometric consistency for large scale image search},
  author={Jegou, Herve and Douze, Matthijs and Schmid, Cordelia},
  booktitle={European conference on computer vision},
  pages={304--317},
  year={2008},
  organization={Springer}
}

Dataset Homepage

Copydays project page