Hugging Face's logo

Datasets:
nvidia
/
miracl-vision

Modalities:
Image
Tabular
Text
Formats:
parquet
Sub-tasks:
document-retrieval
Languages:
Arabic
Bengali
English
Size:
100K - 1M
ArXiv:
Tags:
text
image
Libraries:
Datasets
pandas
License:
Dataset card Data Studio Files Community
1
miracl-vision / README.md
nv-bschifferer's picture
nv-bschifferer
Update README.md
309e169 verified
preview code
|
raw
history blame contribute delete
20 kB
metadata
language:
  - ar
  - bn
  - en
  - es
  - fa
  - fi
  - fr
  - hi
  - id
  - ja
  - ko
  - ru
  - sw
  - te
  - th
  - zh
  - de
  - yo
multilinguality:
  - multilingual
license:
  - cc-by-sa-4.0
task_ids:
  - document-retrieval
tags:
  - text
  - image
configs:
  - config_name: queries-ar
    data_files:
      - split: default
        path: ar/queries.parquet
  - config_name: corpus-ar
    data_files:
      - split: default
        path: ar/corpus.parquet
  - config_name: qrels-ar
    data_files:
      - split: default
        path: ar/qrels.parquet
  - config_name: images-ar
    data_files:
      - split: default
        path: ar/images.parquet
  - config_name: queries-bn
    data_files:
      - split: default
        path: bn/queries.parquet
  - config_name: corpus-bn
    data_files:
      - split: default
        path: bn/corpus.parquet
  - config_name: qrels-bn
    data_files:
      - split: default
        path: bn/qrels.parquet
  - config_name: images-bn
    data_files:
      - split: default
        path: bn/images.parquet
  - config_name: queries-de
    data_files:
      - split: default
        path: de/queries.parquet
  - config_name: corpus-de
    data_files:
      - split: default
        path: de/corpus.parquet
  - config_name: qrels-de
    data_files:
      - split: default
        path: de/qrels.parquet
  - config_name: images-de
    data_files:
      - split: default
        path: de/images.parquet
  - config_name: queries-en
    data_files:
      - split: default
        path: en/queries.parquet
  - config_name: corpus-en
    data_files:
      - split: default
        path: en/corpus.parquet
  - config_name: qrels-en
    data_files:
      - split: default
        path: en/qrels.parquet
  - config_name: images-en
    data_files:
      - split: default
        path: en/images.parquet
  - config_name: queries-es
    data_files:
      - split: default
        path: es/queries.parquet
  - config_name: corpus-es
    data_files:
      - split: default
        path: es/corpus.parquet
  - config_name: qrels-es
    data_files:
      - split: default
        path: es/qrels.parquet
  - config_name: images-es
    data_files:
      - split: default
        path: es/images.parquet
  - config_name: queries-fa
    data_files:
      - split: default
        path: fa/queries.parquet
  - config_name: corpus-fa
    data_files:
      - split: default
        path: fa/corpus.parquet
  - config_name: qrels-fa
    data_files:
      - split: default
        path: fa/qrels.parquet
  - config_name: images-fa
    data_files:
      - split: default
        path: fa/images.parquet
  - config_name: queries-fi
    data_files:
      - split: default
        path: fi/queries.parquet
  - config_name: corpus-fi
    data_files:
      - split: default
        path: fi/corpus.parquet
  - config_name: qrels-fi
    data_files:
      - split: default
        path: fi/qrels.parquet
  - config_name: images-fi
    data_files:
      - split: default
        path: fi/images.parquet
  - config_name: queries-fr
    data_files:
      - split: default
        path: fr/queries.parquet
  - config_name: corpus-fr
    data_files:
      - split: default
        path: fr/corpus.parquet
  - config_name: qrels-fr
    data_files:
      - split: default
        path: fr/qrels.parquet
  - config_name: images-fr
    data_files:
      - split: default
        path: fr/images.parquet
  - config_name: queries-hi
    data_files:
      - split: default
        path: hi/queries.parquet
  - config_name: corpus-hi
    data_files:
      - split: default
        path: hi/corpus.parquet
  - config_name: qrels-hi
    data_files:
      - split: default
        path: hi/qrels.parquet
  - config_name: images-hi
    data_files:
      - split: default
        path: hi/images.parquet
  - config_name: queries-id
    data_files:
      - split: default
        path: id/queries.parquet
  - config_name: corpus-id
    data_files:
      - split: default
        path: id/corpus.parquet
  - config_name: qrels-id
    data_files:
      - split: default
        path: id/qrels.parquet
  - config_name: images-id
    data_files:
      - split: default
        path: id/images.parquet
  - config_name: queries-ja
    data_files:
      - split: default
        path: ja/queries.parquet
  - config_name: corpus-ja
    data_files:
      - split: default
        path: ja/corpus.parquet
  - config_name: qrels-ja
    data_files:
      - split: default
        path: ja/qrels.parquet
  - config_name: images-ja
    data_files:
      - split: default
        path: ja/images.parquet
  - config_name: queries-ko
    data_files:
      - split: default
        path: ko/queries.parquet
  - config_name: corpus-ko
    data_files:
      - split: default
        path: ko/corpus.parquet
  - config_name: qrels-ko
    data_files:
      - split: default
        path: ko/qrels.parquet
  - config_name: images-ko
    data_files:
      - split: default
        path: ko/images.parquet
  - config_name: queries-ru
    data_files:
      - split: default
        path: ru/queries.parquet
  - config_name: corpus-ru
    data_files:
      - split: default
        path: ru/corpus.parquet
  - config_name: qrels-ru
    data_files:
      - split: default
        path: ru/qrels.parquet
  - config_name: images-ru
    data_files:
      - split: default
        path: ru/images.parquet
  - config_name: queries-sw
    data_files:
      - split: default
        path: sw/queries.parquet
  - config_name: corpus-sw
    data_files:
      - split: default
        path: sw/corpus.parquet
  - config_name: qrels-sw
    data_files:
      - split: default
        path: sw/qrels.parquet
  - config_name: images-sw
    data_files:
      - split: default
        path: sw/images.parquet
  - config_name: queries-te
    data_files:
      - split: default
        path: te/queries.parquet
  - config_name: corpus-te
    data_files:
      - split: default
        path: te/corpus.parquet
  - config_name: qrels-te
    data_files:
      - split: default
        path: te/qrels.parquet
  - config_name: images-te
    data_files:
      - split: default
        path: te/images.parquet
  - config_name: queries-th
    data_files:
      - split: default
        path: th/queries.parquet
  - config_name: corpus-th
    data_files:
      - split: default
        path: th/corpus.parquet
  - config_name: qrels-th
    data_files:
      - split: default
        path: th/qrels.parquet
  - config_name: images-th
    data_files:
      - split: default
        path: th/images.parquet
  - config_name: queries-yo
    data_files:
      - split: default
        path: yo/queries.parquet
  - config_name: corpus-yo
    data_files:
      - split: default
        path: yo/corpus.parquet
  - config_name: qrels-yo
    data_files:
      - split: default
        path: yo/qrels.parquet
  - config_name: images-yo
    data_files:
      - split: default
        path: yo/images.parquet
  - config_name: queries-zh
    data_files:
      - split: default
        path: zh/queries.parquet
  - config_name: corpus-zh
    data_files:
      - split: default
        path: zh/corpus.parquet
  - config_name: qrels-zh
    data_files:
      - split: default
        path: zh/qrels.parquet
  - config_name: images-zh
    data_files:
      - split: default
        path: zh/images.parquet

MIRACL-VISION

MIRACL-VISION is a multilingual visual retrieval dataset for 18 different languages. It is an extension of MIRACL, a popular text-only multilingual retrieval dataset. The dataset contains user questions, images of Wikipedia articles and annotations, which article can answer a user question. There are 7898 questions and 338734 images. More details can be found in the paper MIRACL-VISION: A Large, multilingual, visual document retrieval benchmark.

This dataset is ready for commercial usage for evaluation of the multilingual, multimodal retriever pipelines.

Correspondence to

Benedikt Schifferer ([email protected])

Dataset Creation Date:

31st January 2025

License/Terms of Use:

This dataset is licensed under Creative Commons Attribution-ShareAlike 4.0 International. Additional Information: Apache License 2.0.

Intended Usage:

Users can evaluate multilingual, multimodal retriever pipelines.

Dataset Characterization

Dataset Collection Method: Automated Labelling Method: Human

Dataset Format

The images are stored Pillow (PIL) Images in HuggingFace Dataset format The questions, corpus, questions-corpus pairs are stored in parquet/BEIR format

Reference(s): 

@misc{osmulsk2025miraclvisionlargemultilingualvisual,
      title={MIRACL-VISION: A Large, multilingual, visual document retrieval benchmark}, 
      author={Radek Osmulsk and Gabriel de Souza P. Moreira and Ronay Ak and Mengyao Xu and Benedikt Schifferer and Even Oldridge},
      year={2025},
      eprint={2505.11651},
      archivePrefix={arXiv},
      primaryClass={cs.IR},
      url={https://arxiv.org/abs/2505.11651}, 
}

Ethical Considerations:

NVIDIA believes Trustworthy AI is a shared responsibility and we have established policies and practices to enable development for a wide array of AI applications. When downloaded or used in accordance with our terms of service, developers should work with their internal model team to ensure this model meets requirements for the relevant industry and use case and addresses unforeseen product misuse.

Please report security vulnerabilities or NVIDIA AI Concerns here.

Example

The requirements are

pip install qwen_vl_utils beir==2.0.0

The dataset contains an eval_example using MrLight/dse-qwen2-2b-mrl-v1​​

python embedding_eval.py --dataset nvidia/miracl-vision --language en

Loading the Dataset 

from datasets import load_dataset

def hf_beir_queries(queries):
    queries_beir = {}
    for query in queries:
        queries_beir[query['_id']] = query['text']
    return(queries_beir)

def hf_beir_corpus(corpus):
    corpus_beir = {}
    for doc in corpus:
        corpus_beir[doc['_id']] = doc
    return(corpus_beir)

def hf_beir_qrels(qrels):
    qrels_beir = {}
    for el in qrels:
        if str(el['query-id']) in qrels_beir:
            qrels_beir[str(el['query-id'])][str(el['corpus-id'])] = el['score']
        else:
            qrels_beir[str(el['query-id'])] = {str(el['corpus-id']): el['score']}
    return(qrels_beir)

def load_data(
    path,
    lang
):
    queries = load_dataset(path, 'queries-' + str(lang), split='default')
    queries = hf_beir_queries(queries)
    corpus = load_dataset(path, 'corpus-' + str(lang), split='default')
    corpus = hf_beir_corpus(corpus)
    qrels = load_dataset(path, 'qrels-' + str(lang), split='default')
    qrels = hf_beir_qrels(qrels)
    images = load_dataset(path, 'images-' + str(lang), split='default')
    return(queries, corpus, qrels, images)

queries, corpus, qrels, images = load_data('nvidia/miracl-vision', 'en')

Dataset Statistics

Number of Images: 338734

Number of questions: 7898

Total Data Storage: 95GB

MIRACL (original) MIRACL (original) MIRACL-VISION MIRACL-VISION
Language Language Code # of queries # of document chunks # of queries # of documents
Arabic ar 2896 2061414 2127 75444
Bengali bn 411 297265 229 8495
Chinese zh 393 4934368 189 8672
English en 799 32893221 447 42971
Farsi fa 632 2207172 342 15846
Finnish fi 1271 1883509 791 33679
French fr 343 14636953 142 6990
German de 305 15866222 129 6302
Hindi hi 350 506264 184 8004
Indonesian id 960 1446315 603 23842
Japanese ja 860 6953614 387 17909
Korean ko 213 1486752 130 5700
Russian ru 1252 9543918 564 25201
Spanish es 648 10373953 369 17749
Swahili sw 482 131924 239 7166
Telugu te 828 518079 480 15429
Thai th 733 542166 451 16313
Yoruba yo 119 49043 95 3022
Avereage 750 5907342 439 18819

Results

MIRACL-VISION (Text) MIRACL-VISION (Text) MIRACL-VISION (Text) MIRACL-VISION (Text) MIRACL-VISION (Image) MIRACL-VISION (Image) MIRACL-VISION (Image) MIRACL-VISION (Image)
multilingual-e5-large snowflake-arctic-embed-l-v2.0 gte-multilingual-base bge-m3 dse-qwen2-2b-mrl-v1 gme-Qwen2-VL-2B-Instruct vdr-2b-multi-v1 colqwen2-v1.0
LLM Parameters (in M) 560 567 305 567 1543 1543 1543 1543
Language
Arabic 0.8557 0.8754 0.8503 0.8883 0.3893 0.4888 0.4379 0.4129
Bengali 0.8421 0.8325 0.8211 0.8585 0.2352 0.3755 0.2473 0.2888
Chinese 0.6900 0.7179 0.7167 0.7458 0.5962 0.6314 0.5963 0.4926
English 0.7029 0.7437 0.7345 0.7348 0.6605 0.6784 0.6784 0.6417
Farsi 0.6793 0.7001 0.6984 0.7297 0.2250 0.3085 0.2398 0.2616
Finnish 0.8974 0.9014 0.8957 0.9071 0.4162 0.6863 0.5283 0.6604
French 0.7208 0.8236 0.7771 0.8158 0.7160 0.6851 0.7194 0.6876
German 0.7622 0.7774 0.7498 0.7695 0.6267 0.6345 0.6205 0.5995
Hindi 0.7595 0.7255 0.6916 0.7581 0.1740 0.3127 0.2058 0.2209
Indonesian 0.6793 0.6906 0.6757 0.7049 0.4866 0.5416 0.5254 0.5320
Japanese 0.8378 0.8484 0.8442 0.8720 0.6232 0.7305 0.6553 0.6970
Korean 0.7327 0.7545 0.7397 0.7934 0.4446 0.6202 0.4952 0.4419
Russian 0.7857 0.8242 0.8023 0.8363 0.6505 0.7202 0.6995 0.6811
Spanish 0.6596 0.7250 0.7029 0.7268 0.5927 0.6277 0.6274 0.6224
Swahili 0.8157 0.8089 0.7987 0.8337 0.4156 0.5348 0.4509 0.4931
Telugu 0.8948 0.9201 0.9076 0.9090 0.0274 0.0893 0.0318 0.0264
Thai 0.8424 0.8485 0.8509 0.8682 0.2692 0.3563 0.3177 0.2389
Yoruba 0.5655 0.5332 0.5698 0.5842 0.4178 0.4884 0.4577 0.5120
Average 0.7624 0.7806 0.7682 0.7964 0.4426 0.5283 0.4741 0.4728
Average w/o Telugu 0.7546 0.7724 0.7600 0.7898 0.4670 0.5542 0.5002 0.4991